Professional Documents
Culture Documents
Editor
Editorial Board
TZOTZIL CLAUSE
STRUCTURE
1. Introduction 40
2. Moods and Aspects 41
2.1. Neutral Aspect 41
2.2. Incompletive Aspect 41
2.3. Completive Aspect 41
2.4. Perfect Aspect 42
3. Cross-referencing Person 43
3.1. Set A Affixes 43
3.2. Set B Affixes 44
3.3. Imperative Suffixes 45
4. Cross-referencing Number 46
4.1. First Person 46
4.2. Second and Third Person 48
5. The Optionality of Number Agreement 50
6. Agreement and Covert Arguments 53
7. APG Account of Agreement 54
7.1. Agreement Laws 54
7.2. Tzotzil Agreement Rules 57
Notes 59
T ABLE OF CONTENTS IX
1. Introduction 61
2. Syntax of Passive Clauses 61
2.1. Advancement to Subject 64
2.2. Passive Suffixes 65
2.2.1. -e: Monosyllabic Stems 65
2.2.2. -bil: Passive Perfects 66
3. Tzotzil Passive Rules (APG) 66
3.1. Passive Chomeurs 68
3.2. The Form of Chomeurs 68
3.2.1 . Possessor of Relational Noun 69
3.2.2. Object of Preposition 71
3.2.3. Passive Chomeur Rule 72
3.3. Passive Suffixes 73
3.4. Other Passive Rules 74
Notes 74
1. Introduction 77
2. Reflexive Clauses 77
3. Reciprocal Coreference 81
4. Tzotzil Rules (APG) 81
4.1. Lower Pioneer 83
4.2. Conditions on Reflexives 85
Notes 85
1. Introduction 87
2. Reflexive Unaccusative Clauses 88
3. Plain Unaccusative Clauses 91
3.1. Bivalent Stems 94
3.2. Morphological Properties of Bivalent Stems 96
3.2.1. Perfect 96
3.2.2. Subjunctive 97
4. Verb Classification 98
5. Tzotzil Rules (APG) 99
Notes 102
Appendix 103
x TABLE OF CONTENTS
1. Introduction 104
2. Ditransitive Clauses 104
3. 3-to-2 Advancement 106
3.1. Agreement 107
3.2. Ditransitive Passives 108
3.3. Ditransitive Reflexives 110
4. Non-Existence of Final Indirect Objects 114
5. Restrictions on Advancement 114
5.1. Chomage Condition 114
5.2. Person Restriction 116
6. Ditransitive Perfect Passives 117
7. Tzotzil Rules (APG) 118
7.1. 3-to-2 Advancement 118
7.2. Ditransitive Passives 120
7.3. Ditransitive Reflexives 121
8. Conclusion 123
Notes 123
Appendix 124
1. Introduction 126
2. Possessor Ascension 126
2.1. Agreement 130
2.2. Passive 131
2.3. Reflexives 132
3. Coreference Condition 1 134
4. Restriction on Ascension Host 135
5. Tzotzil Possessor Ascension Rule 138
6. The Unique 3 Arc Constraint 138
7. Optional Cases of Possessor Ascension 141
7.1. First and Second Person Possessors 141
7.2. Non-Pronominal Possessors 142
8. Co reference Condition 2 142
9. Possessor Ascension in Discourse 145
10. APG Laws and Tzotzil Rules 147
10.1. Possessor Ascension 147
10.2. Other Rules 151
11. Conclusion 152
Notes 152
Appendix 153
TABLE OF CONTENTS Xl
1. Introduction 155
2. Distinguishing Topic and Focus 157
3. Surface Constituency in Possessor Ascension Structures 160
3.1. Evidence from other Extraction Structures 163
3.2. Other Resolutions of Possessor Ascension 165
4. Topic and Focus 166
4.1. Topicalization 166
4.2. Focus 170
5. Copy and Co referential Pronouns 172
6. APG Laws and Tzotzil Rules 173
6.1. Possessor Ascension and the Successor Erase Law 173
6.2. Surface Constituency in Possessor Ascension Structures 175
6.3. Co reference Rule 175
6.4. Topic and Focus 175
6.4.1. Overlay Relations 175
6.4.2. Tzotzil Constraints 176
6.5. Conclusion 177
Notes 177
1. Introduction 180
2. Possessor Ascension 180
3. Conjunct Union 183
3.1. -chi7uk 183
3.1.1. xchi7uk as Conjunction 184
3.1.2. -chi7uk as Predicate 185
3.1.3. xchi7uk as Flag 186
3.2. Conjunct Union 187
3.3. Reflexive Conjunct Union 190
3.4. Indefinite Comitatives 195
4. Summary 196
5. APG Laws and Tzotzil Rules 197
5.1. Conjunct Union 197
5.2. Conjunct Union Law 199
5.3. Tzotzil Conjunct Union Rules 200
5.4. Reflexive Conjunct Union 202
5.5. Surrogate Agreement 203
6. Conclusion 208
Notes 208
Appendix 210
xii T ABLE OF CONTENTS
1. Introduction 212
2. Causative Clause Union 214
2.1. Evidence for Initial Biclausal Structure 217
2.2. Evidence for Union: Complement Subject 217
2.2.1. Interaction with Passive 217
2.2.2. Interaction with Possessor Ascension 218
2.3. Evidence for Union: Complement Direct Object 220
2.3.1. Chomeur Restriction 221
2.3.2. Reflexive Restriction 223
2.4. Advancements in the Complement 225
2.5. Conclusion 228
3. Abilitative Clause Union 229
3.1. Analysis 231
3.2. Unaccusative Complement 233
3.3. Initial Biclausal Structure 236
3.4. Final Monoclausal Structure 237
3.5. Further Remarks on Inflection 239
3.6. Interaction with Possessor Ascension 242
4. Summary 244
5. APG Laws and Tzotzil Rules 245
Notes 247
Appendix 249
1. Introduction 252
2. Quantifiers 252
3. Prepredicate Quantifiers without Classifier 255
3.1. Monotransitive Clauses 255
3.2. Ditransitive Clauses 255
3.3. Reflexive Clauses 257
4. Prepredicate Quantifiers with Classifier 258
5. Postpredicate Quantifiers 261
6. Grammatical Relations versus Linear Order 264
6.1. Binding Ergatives in Reflexive Clauses 265
6.2. Pima Quantifier Binding 266
7. Conclusion 268
Notes 269
Appendix 269
CONCLUSION 272
TABLE OF CONTENTS XllI
REFERENCES 279
INDEX 282
INTRODUCTION
1. The study of natural language syntax proceeds along two tracks. One
involves the construction of explicit theories which seek to characterize
the class of possible structures of natural language sentences. The other
involves the description of actual syntactic systems. It is generally acknowl-
edged that these activities, though in principle distinct, interact in crucial
ways. In particular, generative linguists agree that the adequacy of a
syntactic theory is determined in part by its success in providing adequate
language-particular grammars.
Given the general agreement on this point, it is surprising how few
generative linguists have attempted to construct consistent, coherent
language-particular grammars within any framework. Notable exceptions
include Dixon (1972), Kayne (1975), and Harris (1981). Of course, none
of these purports to account for all aspects of sentence structure; each is a
grammar fragment. But the domain of each is sufficiently large that serious
problems of consistency arise and have to be resolved. The rules formu-
lated therein claim our serious attention because they cover appropriately
large domains of facts.
What follows is put forth in the same spirit. This book presents an
explicit grammar fragment of Tzotzil, a Mayan language of Mexico - one
which covers enough of Tzotzil clause structure to constitute a reasonable
basis for a complete grammar of the language. The account proposed here
is presented in two ways. Much of the book is devoted to a non-formal
presentation of the analysis, drawing on ideas of relational grammar, as
well as relying on notions of traditional grammar. The analysis is also
formalized within the theory of ARC PAIR GRAMMAR (APG). The choice
of a theoretical framework is a complex one, and in the end, such choices
have to be judged by whether or not they yield adequate and insightful
analyses, and whether or not they raise interesting questions. My choice of
APG was guided by the phenomena I wished to describe, by the generali-
zations that appeared to govern them, and by the desire to give a highly
explicit account. Much of my earlier work on Tzotzil was couched in
terms of RELATIONAL GRAMMAR (RG), but ultimately RG was not a
sufficiently articulated theory to support the kind of analysis developed
here. Unlike RG, APG provides an explicit account of surface structure,
of prepositions, of anaphora, of agreement, all of which figure centrally in
the decription which follows.
The decision to formalize the description in APG terms has posed a
serious problem of exposition since few linguists are familiar with it, and
xv
xvi INTRODUCTION
NOTE
xxi
LIST OF ABBREVIATIONS
In Glosses:
A* Set A* agreement affixes
AI,2,3 Set A agreement affixes - I st person, etc.
BI,2 Set B agreement affixes - I st person, etc.
sg singular
pi plural
plinc plural inclusive
plexc plural exclusive
agn agentive
c1(s) c1itic(s)
comp complementizer
cp completive aspect
dir directional
icp incompletive aspect
imp imperative
io indirect object
nc numeral classifier
nt neutral aspect
pf perfect
poss inanimate possession
ppf passive perfect
psv passive
subj subjunctive
topic topic flag
3 existential predicate
? Q'c1itic
emphatic particle
Others:
APG Arc Pair Grammar
J&P Johnson and Postal (1980)
PA Possessor Ascension
PN Pair Network
PTAC Potential Tzotzil Agreement Controller
RG Relational Grammar
TG Transformational Grammar
I Subject
2 Direct Object
3 Indirect Object
32A 3-to-2 Advancement
iff if and only if
xxiii
GRAMMATICAL NOTES
1. INTRODUCTION
Many of the examples cited in later chapters were culled from texts, and
are 'uncontrolled' in the sense that they contain grammatical features
irrelevant to the point at hand. This is distracting at times, but I have left
them intact, for they offer some sense of a Tzotzil utterance, and may be
stimulating in unforeseen ways. This chapter aims to provide an informal
sketch of Tzotzil syntax and inflectional morphology so that the reader
may successfully manoeuvre such examples. The sketch makes no claims
to completeness or rigor. The choice of material was determined by the
grammatical features of examples actually cited. For more comprehensive
treatments of Tzotzil syntax, the reader should consult the works cited in
the introduction.
2. BASICS
(1) illustrates basic VOS order, as well as the fact that neither subject nor
object is case-marked. The verb bears a prefix, glossed 'A3', which cross-
references the 3rd person transitive subject. The direct object, being 3rd
person, is marked by no overt affix. Lok'el is a so-called 'directional' (see
below); the last nominal in the sentence is marked for its definiteness by
the article ti, which generally cooccurs with the enclitic -e, as it does here.
Overt agreement with the object is illustrated in (2):
The prefix glossed 'B l' cross-references the person of the direct object;
the suffix glossed '1 plinc' (1 st person plural inclusive) cross-references its
number (and redundantly its person). Example (2) consists of only a verb,
but is a perfectly well-formed sentence. Its nominal arguments, being
personal pronouns, are not pronounced.
Examples (3) and (4), together with (1) and (2), suggest the ergativity
of the agreement system. In (1), the 3rd person direct object is cross-
referenced by 0. 0 also cross-references 3rd person subjects in intransi-
tive clauses:
(3) 7i- tal.
cp come
He/she/it/they came.
Example (2) shows that a 1st person plural inclusive direct object is cross-
referenced by the prefix -i- and the suffix -otik. The same affixes cross-
reference 1st person plural inclusive subjects in intransitive clauses:
(4) L- i- tal -otik.
cp B1 come 1plinc
We (inclusive) came.
The agreement system formally equates subjects in intransitive clauses
and direct objects in transitive clauses, cross-referencing both by the same
set of affixes. These two relations together constitute the ABSOLUTIVE
relation. The affixes which cross-reference absolutives are termed SET B by
Mayanists. Subjects of transitive clauses (ERGATIVES) are cross-referenced
by a different set of affixes which Mayanists call SET A. For example, 3rd
person ergatives are cross-referenced not by 0, but by S-, as in (1). 1st
person plural inclusive ergatives are cross-referenced not by -i- •.. -otik,
but by j- ••. -tik:
(5) 7i- j- pet -tik lok'el ti vinik -e.
cp A1 carry 1plinc away the man c/
We (inclusive) carried away the man.
(In the glosses, 'AI' = Set A, 1st person, 'B2' = Set B, 2nd person, etc.).
of) predicates. The examples cited above contain V predicates, but Ns and
As also head phrases in predicate function:
(6) 7antz -on.
woman Blsg
I'm a woman.
3.1. V
Only V stems inflect for aspect (neutral (nt), completive (cp), incompletive
(icp), perfect (pf)). The two principal subclasses of V are transitive and
intransitive verbs - classes sharply distinguished by their inflection (see
chapter 3).
3.2. N
b. s- na Ii Maruch -e
A3 house the Maruch cl
Maruch's house
c. s- tzek Ii 7antz -e
A3 skirt the woman cl
the woman's skirt
with totil 'father', and tzekil 'skirt' which cannot be. 2 (See chapter 8, and
Haviland (1981, chapter 7.6).)
S modifiers are relative clauses. In some cases, these follow the head
directly, while in others, they are introduced by ti or Ii, which here
apparently function as complementizers.
(15) a. ... ti tak'in 7i- s- bik' 7une OCK 178
the money cp A3 swallow cls
the money that he had swallowed
3.3. A
The A class includes all stems which can function as predicates, but which
are neither verbs nor nouns. Unlike verbs, As do not inflect for aspect.
Unlike nouns, they do not head NPs, and in particular, cannot combine
with genitives. Examples:
(16) Tzotz -on.
strong Blsg
I am strong.
(17) Mas bik'it -oxuk. OCK 46
more little B2pl
You (pi) are even smaller.
(18) 7ip xa 70nox 7un. OCK 195
sick cls
She was still sick.
Some A stems can function both as predicates and as N modifiers
(i.e., attributively), e.g., 7ach' 'new', 7unen 'small', and yan 'other', and
the Spanish loanwords sonso 'foolish' and prove 'poor'. Some A stems
function predicatively but not attributively. These include the existential
predicates (see below), A stems which describe positions, like chotol
'seated' and va7al 'prone', as well as yox 'green', sak 'white', muk' 'large',
7ep 'much', takin 'dry', nat 'deep, long, tall'. Some of these are deriva-
tionally related to stems which function attributively and not predicatively,
6 CHAPTER I
e.g., yaxal 'green', sakil 'white', muk'ta 'large', 7epal 'many', taki 'dry', natil
'deep, long, tall'. It is not clear what lexical class the latter stems belong to:
if they are A stems, then that class must contain a subclass whose
members cannot function predicatively.J
3.3.l. Quantifiers
Tzotzil has a set of number roots (e.g., cha7 'two', chan 'four') which do
not themselves function as stems, but constitute the base for processes
which derive stems of various classes (Fleck 1981). Numbers which are
used both attributively and predicatively to quantify over NPs (i.e., to
count the items denoted by the NPs) are formed by compounding a
number root with a NUMERICAL CLASSIFIER. The classifier restricts the
class of items being counted, usually in terms of some salient physical
property (cf. Berlin 1968). (This seems analogous to restricted quantifica-
tion in logic.) V07 is used for counting humans, kot for quadrupeds, p'ej
for squattish things, and so on. The resulting stem is an A stem, and can
function as a modifier (19)-(20) or as a predicate (21 )-(22):
(19) ti chan -v07 kremotike
the four nc boys
the four boys
b. Te -ot.
there B2sg
You're there.
Te and 1i7 also function as the heads of adverbial phrases:
(27) 70y k'in te ta lobel.
3 festival there in SC
There is a festival in San Cristobal.
See also section 7.4.
4.1. Clitics
4.1.1. Clause-proclitic
4.1.2. S-enclitic
There are two S-final clitics: e and 7un. 7un cliticizes to any syntactic
category. In (29a), it cliticizes to a particle. It occurs twice in (29b) - once
cliticized to a verb and once to a verb + clitic. In (31), it cliticizes to a
noun:
(29) a. Mi lasut tal noxtok 7un?
? you return here again cl
Have you come back again?
4.1.3. V-enclitic
4.1.4. Clause-second
This class is the largest and contains temporal! aspectual (t/a), modal, and
evidential clitics. Its most common members are listed in (34) (glosses are
approximate ).
(34) T/A MODAL
xa 'now, already' nan 'maybe'
to 'still' kik'maybe'
70x [completed time] me 'please'
no 'just, simply'
EVIDENTIAL
la [quotative, on authority of someone other than the speaker]
Very crudely, these clitics occur in clause-second position. In clauses
consisting of a one-word predicate plus nominal arguments, they im-
mediately follow the predicate:
(35) 7i- bat xa Ii Xun -e.
cp go cl the Xun cl
Xun has already gone.
4.2. Directionals
Directionals are derived from a closed class of intransitive verbs by
suffixation of -el. All members of this class except one are monosyllabic,
and all except one denote motion or its absence. Directionals follow the
predicate directly (separated only by clause-second clitics and the reflexive
10 CHAPTER 1
nominal (on the latter, see chapters 5, 7, 8» and usually denote the
direction or trajectory of an action.
(38) ROOT DIRECTIONAL
bat 'go' batel 'from time to time'
7ech' 'pass by' ech'el 'away'
7och 'enter' ochel'in'
jelav 'pass by, reach' jelavel 'through, by'
kom'remain' komel 'remaining'
k'ot 'arrive' k'otel 'arriving'
10k' 'leave' lok'el 'out'
muy 'ascend' muyel 'up'
sut 'return' sutel 'back'
tal 'come' tal/talel 'here'
vay'sleep' vayel 'sleeping'
yal 'descend' yalel 'down'
Examples (see also (1 »:
(39) 7ich' -0 ech'el. OCK 350
take imp away
Take it away!
4.3. Particles
5. FLAGGING
6. WORD ORDER
7. CONSTRUCTION SURVEY
7.1. Negation
Sentences are negated by combining mu or muk' (or an elaboration of one
GRAMMATICAL NOTES 13
of these words, e.g., muk' bu 'never') with a clause. The negative word
precedes the clause. When mu combines with a clause whose predicate
is nonverbal, the predicate is suffixed with -uk (-ik- word-internally).
Compare the negation of nominal and adjectival predicates with the
negation of a verbal predicate:
"mu generally contrasts with muk', implying speaker's unwillingness to carry out action;
mu xibat. 'I will not go.' muk' xibat 'I am not going.' When occurring with interrogative
particle, mu implies speaker's desire that person addressed carry out action; mi mu xabat.
'Won't you go?' mi muk' xabat? 'Aren't you going?'"
7.2. Questions
In the second set, the predicate bears no affix, and the argument
corresponding to its object is understood as coreferent with the subject of
the main clause. These are analogous to Equi or Raising constructions.
(70) 7a Ii chon -e, mu s- k'an mil -el. OCK 166
topic the snake cl not A3 want kill el
The snake doesn't want to be killed.
Notes:
7.5. Topics
NOTES
I J-jteklum-e bears the only derivational prefix in the language. J- forms agentive nouns,
1. ARCS
TABLE I
Subject I
Direct Object 2
Indirect Object 3
Chomeur Cho
Genitive Gen
Head H
Predicate P
Union U
Dead Head
Conjunct Con
Locative Loc
Benefactive Ben
Instrument Inst
Clitic CI
20
THEORETICAL SKETCH 21
Table I (Continued)
Affix Af
Stem St
Flag F
Marquee Marq
Topic Top
Question Q
Focus Foc
L Label
GRx C"," +1
b
Here, the node labelled b bears a grammatical relation to the node
labelled a. Node b is A's HEAD, a is A's TAIL. Node a GOVERNS band,
conversely, b is a DEPENDENT of a. The name of the relation b bears to a
is written to the left of the arrow. The coordinate sequence naming the
levels in which the relation is borne is written to the right.
The sentence Sally visited Marta involves three linguistic states, hence
three arcs:
(2) 80
The arc labelled A represents the fact that Sally bears the subject relation
in its clause at the c l level. Formally, the node associated with Sally is A's
head, and the node labelled 80 is its tail. l Arc B represents the fact that
22 CHAPTER 2
visited is the predicate of the same clause at the same level, and C the fact
that Marta is the direct object in the same clause at the same level.
The sentence Marta was visited by Sally involves at least five linguistic
states, hence at least five arcs (I ignore the auxiliary was and the preposi-
tion by). Its representation contains both nominal arcs (A and C) in
example (2), but contains in addition 0 and E. Arc 0 represents the fact
that Sally bears the chomeur relation in the clause at a level distinct from
that at which it bears the subject relation, and Arc E represents the fact
that Marta bears the subject relation in the clause at the same level that
Sally bears the chomeur relation.
(3)
80
Cho c 2
R-SIGNS
Structural RS
Nominal RS
Cho Dead
Nuclear Term RS
1.2. Stratum
The notion Ckth stratum of b is defined as the maximal set of arcs with tail
b having coordinate Ck' This allows definitions of the terms INITIAL and
FINAL strata. The initial stratum of b is the set of arcs with tail b having
the C 1 coordinate. Stratum Ck is the final stratum of b if and only if there is
no cj stratum of b, j > k.
Now it is possible to define TRANSITIVE stratum and INTRANSITIVE
stratum. The definition of the first is straightforward:
(7) Def: cn is a TRANSITIVE STRATUM iff Cn contains a 1 arc and a
2 arc.
RG recognizes two classes of intransitive strata: those containing a 1 arc
and no 2 arc (roughly the traditional definition), and those containing a 2
arc and no 1 arc. The former are UNERGATIVE strata, the latter UNACCU-
SATIVE strata. INTRANSITIVE strata are those containing one and only one
nuclear term arc.
(8) Def: Cn is an INTRANSITIVE STRATUM iff clI contains one and
only one nuclear term arc.
Neighbor (A, B)
Overlap (A, B)
Two arcs OVERLAP if they share a head. They are NEIGHBORS if they
share a tail. Arcs which share both a head and a tail are PARALLEL. Arcs
in the KISSES relation share a head but have distinct tails. Arcs in the
COLIMBS relation share a tail but have distinct heads. One arc SUPPORTS
another when the head of the first is the tail of the second. Conversely, the
second is a BRANCH of the first.
26 CHAPTER 2
®~====I®
2.1. Successors
P Cho
seems
Jo be happy
2.2. Replacers
The Replace relation is central to the account of pronouns and flagging
structures. As an illustration of how pronouns are handled, consider the
sentence Maggie said that she was there under the coreferential interpreta-
tion. Maggie heads initial 1 arcs in both the complement clause and the
main clause (B and C). However, Maggie does not head the complement
1 arc in surface structure, this arc being headed by the pronoun she.
Accordingly, A, an arc headed by she, replaces the complement 1 arc, B,
and erases it (A is a pronominal arc (p. 81) and an anaphoric arc (p. 82)):
THEORETICAL SKETCH 29
(19)
1 2
said
the heads of replacer arcs in APG.3 Since such pronouns are not in the
initial stratum, one might wonder how this is to be reconciled with
arguments articulated within TG (e.g., Hankamer and Sag (1976» that at
least some pronouns are 'deep'. The answer is provided in part by the
Coreferential Arc Law (PN Law 90) which requires that any arc headed
by a coreferential pronoun replace a c, arc. As a consequence, nothing
that would correspond to a derived constituent or a moved NP in TG can
be replaced by a pronoun (see J & P, section 11.6).
See chapter 4, section 2.2 for a discussion of the role of replacer arcs in
flagging structures.
Replacers always erase the arcs they replace:
(22) PN Law 1 (Replacer Erase Law):
If A replaces B, then A erases B.
We can now return to the conditions under which successors erase their
predecessors. A successor erases its predecessor unless the latter has a
replacer. That is, there are cases in which an arc has both a successor and
a replacer. In such cases, the replacer erases the arc, and it is only in such
cases that a successor fails to erase its predecessor:
(23) PN Law 2 (Successor Erase Law):4
If A is B's successor and B has no replacer, then A erases B.
The formulation of the Successor Erase Law guarantees for certain
cases that an arc will have only one eraser. This is consistent with a further
law which requires that no arc have more than a single eraser:
(24) PN Law 3 (The Unique Eraser Law):
If A erases Band C erases B, then A = C.
The formulation of the Successor Erase Law guarantees this by giving
replacers precedence over successors in erasure of predecessors. 5
3. ANCESTRAL RELATIONS
By prefixing any relation with 'R-' (for Remote-), one can designate the
ancestral of any relation (see Postal 1986a, p. 34). This yields a set of arcs
each of which bears the relation in question to some other arc in the set.
For example, if A is B's R-predecessor, then A is B's predecessor or A is
the predecessor of some predecessor of B or A is the predecessor of some
predecessor of a predecessor of B, etc. In addition, R-relations are
reflexive, with the consequence that while, for example, no arc can be its
own predecessor, every arc is its own R-predecessor.
To illustrate, consider (25), the (partial) PN for Marta was given the
message by Sally:
(25)
4. PAIR NETWORKS
The PN of (26) is: IHA,Al IB,Bl ID,Dl IF,Fl !B,q ID,Ell, He,Bl,
IE, D HI. Its R -graph is: IA, B, e, D, E, F l, the set of all arcs in (26). Its
L-graph is: IA, B, D, F l, the set of all self-sponsoring arcs. Its S-graph is:
IA, e, E, F l, the set of all unerased arcs.
The Successor Erase Law and the Replacer Erase Law are motivated in
THEORETICAL SKETCH 33
part by the condition that no surface arcs may overlap. This condition is
part of the definition of S-graph, which requires that S-graphs be trees.
Successor/predecessor pairs always overlap, so erasure under the Suc-
cessor Erase Law always resolves overlap. In cases where the cosponsors
of a replacer arc overlap, as in cases involving pronominalization, erasure
under the Replacer Erase Law resolves overlap.1>
6. COORDINATE DETERMINATION
(28)
seems
Jo be happy
Jo seems to be happy.
The first coordinate associated with a replacer arc is 1 greater than the
last coordinate associated with the replaced arc (PN Law 23). Example
(19) is repeated as (29) with lawful coordinates added.
(29)
said
P
contained in the immediately preceding stratum. This rules out PNs like
(30) which contain 'excrescent' strata:
(30)
Sally visited
\
Marta
The Fall-through Law (PN Law 26) guarantees that an arc will have the
right number of coordinates, neither too many nor too few. In essence, it
says that if an arc A occurs in one stratum of b, then it occurs in the next
unless it is erased by some distinct neighboring arc which is in the next
stratum. This requires that an arc fall through to the next stratum in
certain circumstances, and prohibits fall-through when those conditions do
not hold.
The Fall-through Law has two important consequences. One is that any
arc which self-erases is in the final stratum associated with its tail (i.e., is a
final arc). This follows because such arcs do not have distinct local erasers
(recall that no arc has more than one eraser). Therefore, a self-erasing arc
must fall through to the last motivated stratum associated with its tail, that
is, the final stratum. Any arc which has a foreign eraser is also a final arc.
Such an arc lacks a local eraser, and must therefore fall through to the last
motivated stratum associated with its tail, i.e., the final stratum.
8. WORD ORDER
J & P present three possible APG accounts of word order, and Postal
(1986) presents a fourth. All four approaches attempt to reconstruct the
intuition, due to RG, that word order is relevant only at superficial levels
of structure. However word order is described, the simplest approach in
APG will be one in which only swface arcs (unerased arcs) are involved.
Whether this is empirically correct is discussed at some length in J & P
(chapter 12). While no formal account of word order will be given here,
all informal statements about word order refer to the heads of surface
arcs, and only these.
(33)
UN a
38 CHAPTER 2
Accordingly, the APG Chomeur Law (PN Law 62) requires roughly
that an overrun arc have a Cho arc successor only if it would otherwise
fall through into the next stratum. As long as it is erased by a neighboring
arc, it will not. The eraser may be the successor of the arc in question, or,
in certain cases, the overrunner itself.
The following statement of the Chomeur Law incorporates both the RG
Motivated Chomage Law, and the insight just mentioned. R An EMPLOYED
arc is any nominal arc which is not a Cho arc.
(34) Chomeur Law:
B has a Cho arc successor D whose first coordinate is Ck iff
there is some A which overruns B in Ck and there IS no
employed arc C, C i- B, such that C local erases B.
NOTES
I Technically. the head node is also named by an integer, like the tail node. In APG, Sally
J Rosen (1981, chapter 4) takes a different view, arguing from Italian facts that some
anaphoric pronouns head initial arcs. These facts may require a revision of the general
APG position but do not (a priori) undermine such an analysis in particular cases.
4 A revision of this law is proposed in chapter 9.
5 The principal motivations for giving replacers precedence over successors are these: it
allows an elegant reconstruction in terms of erasure of Ross's (1967) observation that
"chopping" constructions, but not "copy" constructions, obey island constraints (J & P, p.
194, fn. 3). Further, it allows for a maximally simple account of coordinate assignment (see
J & P, p. 177). J & P also claim as relevant (p. 518) the fact that this assumption makes
possible simpler accounts of the difference between copy and non-copy structures in
particular grammars. Here the savings seem negligible.
6 The requirement that an S-graph be a tree, and hence contain no overlapping arcs,
cannot be satisfied by self-erasure of one or both of the arcs in question. This is guaranteed
by the Internal Survivor Law (J & P, p. 526).
7 Harris's (1984) analysis of the Georgian facts does not, however, involve an overrun I
arc. Rather, the inital I demotes to 3 in the second stratum, and the 2 advances to I in the
third. My proposal is motivated by certain facts of agreement (see Aissen, to appear). Fur-
ther, I see no motivation for the additional stratum Harris posits.
8 J & P discuss alternatives (p. 356).
'--Clt-\r 1 CI\. .J
1. INTRODUCTION
'B l' cross-references the first person 2, 'A3' the third person 1, and 'A2'
the second person genitive. None of the cross-referenced nominals appear
because they are all non-emphatic pronouns.
Tzotzil agreement involves two categories: person, and number. Agree-
ment in person is obligatory, while agreement in number is optional, at
least from a strictly clause-bounded point of view. From a broader point
of view, whether or not the distinction between singular and plural is made
depends in part on the person/animacy hierarchy. First and second person
nominals always determine number agreement somewhere in the sentence,
while third person nominals do not. This is taken up in sections 4 and 5.
The form agreement actually takes in particular cases is complicated by
the fact that both set A and B affixes come in two subsets. Which form of
a set A (ergative/genit~e)affi~is used depends on the initial segment of
the stem. Which set B. (absolutive) form is used depends chiefly on the
aspect of the predicate. Sections 2 and 3 are devoted to these details.
Section 6 presents an APG theoretical framework for the description of
agreement. There, a number of laws are proposed as universal conditions
on agreement. The Tzotzil agreement rules established account not only
for agreement in simple (monostratal) clauses, but for agreement in all
the ditransitive constructions discussed in subsequent chapters, with the
exception of two cases of plural agreement. Those involve extensions of
the theory of agreement proposed here, and are discussed in chapter 10.
40
INFLECTION AND AGREEMENT 41
Neutral aspect (nt) is marked by the prefix x-. It is used principally after
negation:
(2) Mu x- bat.
not nt go
He/she/it/they isn't/aren't going.
The prefix x- is subject to several morphophonological processes. See
Phonological Rules 3,4, 5.
TABLE II
Summary of Moods and Aspects
INDICATIVE
Aspects:
neutral (nt) x+V
incompletive (icp) ta x + V
completive (cp) I+V With set B prefixes (-i. -a) only
0+V With A2 prefix (a-/av-) only
7i+V Elsewhere
perfect (pf) V;+em Intransitive verbs only
VI +oj Transitive verbs only; active perfect
VI + bi! Transitive verbs only; passive perfect
IMPERATIVE X,+an Intransitive stem only
XI+o Transitive stem only
SUBJUNCTIVE Xi +uk Intransitive stem only
-uk > -ik- word-internally
imperative suffix -an substitutes for
expected -ik-ot in 2nd person sg
imperative suffix -ik substitutes for
expected -ik-oxuk in 2nd person pi
XI Transitive stem only
INFLECTION AND AGREEMENT 43
3. CROSS-REFERENCING PERSON
Set A affixes prefix directly onto the stem. There are two variants dis-
tributed according to the initial segment of the stem. All verb stems have
an underlying initial consonant. Stems with initial glottal stop (7) take one
variant, all other stems take the other. However, 7- drops after a set A
prefix with the result that superficially, one variant (prevocalic) always
appears immediately before a vowel, and the other (preconsonantal)
before a consonant (see Phonological Rules 1).
(7) Examples:
Verbs:
-7iI 'see'
k-iI-oj 'I/we have seen it.'
av-i1-oj 'You have seen it.'
y-il-oj 'He/she/they have seen it.'
-man 'buy'
j-man-oj 'I/we have bought it.'
a-man-oj 'You have bought it.'
s-man-oJ 'He/she/they have bought it.'
Nouns:
-7ixlel 'younger sister'
k-ixlel 'my/our younger sister'
av-ixlel 'your younger sister'
y-ixlel 'his/her/their younger sister'
44 CHAPTER 3
There are two subsets of set B affixes: one is a prefix set, the other a suffix
set.
(8) Set B Affixes
SUFFIX PREFIX
Blsg -on Bl -1-
b. Mu x- i- bat.
not nt Bl go
I won't go.
c. Ta x- i- bat.
icp Bf go
I'm going.
b. *Ch- i- a- mil.
icp B1 A2 kill
The chart in (8) reveals an important structural difference between the
prefix set and the suffix set. Like set A prefixes, set B prefixes mark only
person, while set B suffixes conflate person and number. The result is that
a verb with the B 1 prefix -i-, for example, may have either a 1st person
singular or plural 1, while a verb with the suffix -on 'Blsg' can only have a
1st person singular 1 (see section 5 below).
The form of the imperative suffix depends on the transitivity of the stem.
Transitive stems add -0. Intransitive stems add -an:
(11) a. Man -o!
buy imp
Buy it.
b. Bat -an!
go imp
Go!
*Man-an! and *Bat-o! are ill-formed. The suffix -an is analyzed here as a
set B affix, and -0 as a set A affix. This will account for the fact that the
former occurs only on intransitive stems, and the latter only on transitive
ones, and will simplify the agreement rules. 2
Non-verbal predicates form the imperative in the same way:
4. CROSS-REFERENCING NUMBER
The affixes which mark number are most easily discussed by person.
The form of the 1st person plural suffix depends on two things. One is
whether the hearer is included in the reference of the cross-referenced
nominal. If so, the inclusive form is used. The exclusive form is used
otherwise. The other is the grammatical relation of the cross-referenced
nominal. The plurality of final ergatives and genitives is marked by the
suffixes -tik (inclusive) and -tikotik (exclusive). I call these set A * suffixes:
(16) Examples:
Verbs:
k-il-oj-tik 'We (inc) have seen it/him/them.'
k-il-oj-tikotik 'We (exc) have seen it/him/them.'
j-man-oj-tik 'We (inc) have bought it.'
j-man-oj-tikotik 'We (exc) have bought it'.
Nouns:
k-ixlel-tik 'our (inc) younger sister'
k-ixlel-tikotik 'our (exc) younger sister'
j-tot-tik 'our (inc) father'
j-tot-tikotik 'our (exc) father'
The plurality of first person absolutives is marked by set B suffixes. In
those contexts where set B prefixes are required (i.e., on forms with an
aspectual prefix), the suffix cooccurs with the appropriate set B prefix:
(17) a. Ch- i- tal -otik.
icp B I come B Iplinc
We (inc) are coming.
marked only by the suffix. In Zinacantec Tzotzil, the set B prefix and
suffix generally cooccur only when the suffix is plural. A word with a set B
prefix and a singular set B suffix is ill-formed: *Ch-i-tal-on icp Bl come
Blsg 'I am coming', *Ch-i-s-mil-on icp B1 A3 kill B1sg 'He's going to kill
me.' 3
The set B 2nd person suffix -oxuk cross-references plural 2nd person
nominals wherever set B suffixes are required (on all forms lacking aspect
prefixes):
Otherwise, the plurality of all 2nd and 3rd person final ergatives,
absolutives, and genitives is cross-referenced by -ik. The suffix -ik cross-
references 2nd and 3rd person ergatives in (21), 2nd and 3rd person
possessors in (22), intransitive Is in (23), and 2s in (24).
b. Mi a- man -ik?
? A2 buy 2p/
Did you (pi) buy it?
b. a- tot -ik
A2 father 2p/
your (pi) father
8
prefIX I -i- Person of final absolutives Only on predicates with
2 -a- aspect prefixes (neutral,
incompletive, completive)
Although two factors obscure it, the general principle is that agreement in
number is optional. Since agreement in person is obligatory, this entails
that person agreement and number agreement are distinct. Number
agreement is optional in the sense that when the 1 or 2 of a predicate
refers to a plural entity, an affix cross-referencing the plurality of that
nominal is optional. An example like (25) in which the 1 is plural and the
verb unmarked for plurality illustrates this clearly:
(25) Ch- bat ti ninya- etik -e. OCK 208
icp go the girl pi cl
The girls are going.
Consider also the following which contains two clauses, each with the
same 3rd person plural pronominal 1. The verb of the first clause is not
marked for plurality while the verb in the second is. The (b) example
shows that the intransitive verb ve7 can be inflected for a plural 1:
(26) a. Ba7yi ch- ve7, ba7yi ch- [y]- uch' -ik v07. OCK 204
first icp eat first icp A3 drink 3pl water
They ate first, they drank water first.
In spite of this, I claim that number agreement is optional with all persons,
for examples like the following make it clear that the number of a 1st or
2nd person plural 1 or 2 does not have to be marked on the predicate:
(28) a. Mu j- k'an x- i- bat -otikotik 7une. SSS 92
not A I want nt B I go B Iplexc cls
We didn't want to go.
b. 7i- j- k'exta j- k'u7 -tikotik ta be. SSS 78
cp Al change Al shirt A *Ip/exc on road
We changed our clothes on the road.
(29) Ja7 xa ch- a- sa7 a- tak'in -ik. OCK 197
cl icp A2 seek A2 money 2pl
You (pi) will seek your money.
It is characteristic of such examples that the nominal in question bears
more than one grammatical relation, and controls number agreement
somewhere else in the sentence. In (28b) and (29), the 1st person pronoun
is both 1 and possessor of the 2, and controls number agreement as
possessor. Overt plural agreement with the possessor obviates the need to
mark plurality on the predicate. Note that it does not prevent plural
marking on the predicate:
(30) 7i- j- lap -tikotik Ii j- batz'i k'u7 -tikotik. SSS 78
cp A I wear A *I plexc the A I real clothes A *Iplexc
We put on our native dress.
In (28a), the 1st person pronoun is 1 in both the main and complement
clauses. The complement predicate is marked for the plurality of its 1,
making plural agreement on the main predicate unnecessary.
The difference between 3rd person and 1st12nd person is not that
number agreement is optional with the former and obligatory with the
latter, but that the plurality of a 1st or 2nd person nominal must appar-
ently be cross-referenced somewhere in the sentence, while that of a 3rd
person nominal can be established somewhere else in the discourse, or
pragmatically.
Something must account for the difficulty (or impossibility, perhaps) of
interpreting the 1 in relevant parts of (27) as plural, but making plural
agreement obligatory will incorrectly rule out examples like (28)-(29).
Hence, I assume that number agreement is optional, and that some other
principles, which distinguish 3rd person from 1st and 2nd, account for the
need to mark the plurality of the latter somewhere in the sentence.
There is another factor which obscures the optionality of number
agreement. It is possible for number agreement to be optional and person
agreement obligatory only to the extent that the affixes which mark person
and number are distinct. Recall that set B suffixes conflate person and
52 CHAPTER 3
Af St
TABLE IV
Grammatical Terminals for Agreement Arcs
j-/k- A I
a-lav- A 2
s-/y- A 3
-tik A* plinc
-tikotik A* I plexc
0 B 3
-on B,uf I sg
-ot B,uf 2 sg
-otik B,uf plinc
-otikotik B,uf I plexc
-oxuk B,uf 2 pi
-1- Bpre I
-a- Bpre 2
-ik pi 2or3 pi
-an B 2 imp
-0 A 2 imp
Arcs A and B are headed by agreement affixes. Arc A's head cross-
references the nominal which heads E, and B's head cross-references the
nominal which heads D. Node 100 is the tail of both A and B because the
verb it represents is predicate in the clause containing the cross-referenced
nominals. The natural approach to agreement within APG is one in which
the arc headed by the cross-referenced nominal sponsors the arc headed
by the agreement affix, for this represents the fact that the presence of the
affix depends on the presence of the controller. Agreement affixes head
AGREEMENT ARCS, a term which will remain undefined. Controllers head
CONTROLLER ARCS. Under this approach, E will sponsor A and D will
sponsor B, thus establishing a formal connection between the heads of E
and A, and those of D and B. Y Such a relation is desirable, for both
agreement laws and rules need to refer to it.
There are a number of generalizations about agreement controllers
which, if true, should be expressed as laws. Such laws make substantive
universal claims about agreement, and at the same time, given the APG
conception of the relation between (universal) laws and (language-specific)
rules, allow simplification of Tzotzil rules, since the latter need to state
only what is peculiar to Tzotzii. The laws which follow contain undefined
terms, so they should be regarded as suggestive.
The first is that agreement controllers always head final arcs. A law
requiring this, as in (38), accounts for the fact that in languages with
subject agreement, it is the superficial subject which controls agreement,
and not an earlier subject. (This is controversial, of course; see chapter 10
for discussion.)
(38) Controller Agreement Law
If A sponsors an agreement arc then A is in the final stratum.
The second law concerns the relation between a controller and the
element which bears the agreement affix, e.g., the verb in (37). In general,
agreement is between elements within the same relevant domain: a
predicate agrees with the 1 of its own clause, a possessed noun agrees
with its own possessor. I 0 In present terms, what is at issue is the relation
between the controller arc and the relevant R-support of the agreement
arc. (Recall (chapter 2, section 3) that a R(emote)-support may (directly)
support its branch, or it may support a support of its branch, etc. See
footnote 11 below on this distinction.) The crucial point is that they must
be neighbors. In the case of predicate agreement that P arc which
R-supports the agreement arc must be a neighbor of the controller arc. I I
In the case of genitive agreement, that H arc which R-supports the agree-
ment arc must be a neighbor of the controller arc.
Certain general principles seem to determine the R-sign of the relevant
R-support of an agreement arc. Agreement with clausal dependents is
marked on the predicate. Agreement with nominal dependents is marked
INFLECTION AND AGREEMENT 57
NOTES
I There is some evidence that perfects are not verbs, but A's, for they suffix -uk/-ik- under
negation (see chapter 1, section 7.1): Mu k-iI-oj-uk, mu k'eI-oj-uk. (OCK 128) 'I didn't see
it, I wasn't watching'. Mu ve7-cm-ik-on. 'I haven't eaten'. Mu meIuan-bil-uk ta j7alvanil.
(OCK 237) 'It wasn't built by masons'. If so, perfect suffixes change word class member-
ship. Cf. note 7.
2 The suffix -an will never cross-reference a 2, because no imperative can have a final 2nd
person 2: the I of an imperative is always 2nd person, and will therefore be coreferential
with a 2 of the same person. Any 2s which are coreferential with the I yield reflexive
pronouns which are syntactically 3rd person. See chapter 5.
1 In some dialects of Tzotzil, a set B prefix and singular set B suffix do cooccur. Huistec
Tzotzil requires both when the cross-referenced nominal is the 2 of the clause, though not
when it is intransitive 1 (Cowan 1969). John Haviland (personal communication) informs
me that Zinacantec speakers sometimes also use prefix and (singular) suffix when the
absolutive is direct object.
4 The verb bat 'go' idiosyncratically takes -tik rather than -otik as the B I plinc suffix; two
verbs here are suffixed with the c1itic -e. Note that this text reports past events involving
both the narrator (speaker) and the addressee (hearer); for that reason inclusive plural
forms are used.
, In the 3rd person, person and number are not conflated. I posit a set B affix 0 which
marks 3rd person, but is unmarked for number. The suffix -ik is not a member of set B.
Hence, predicates like the following may have either singular or plural Is:
(i) Kapem.
angry
She/ he/ it/ they is/ are angry.
that perfects (at least intransitive perfects) are non-verbal. Cf. note 1.
H It is not the case, however, that all covert arguments are cross-referenced through
agreement. In ditransitive clauses, the final chomeur may drop when it is pronominal, but it
does not control agreement. Its presence is deducible from the form of the verb, but not
through agreement morphology. See chapter 7.
9 These remarks concern the nature of agreement control, but not the nature of agreement
itself. Agreement has to do with the compatibility of categories associated with the
controller and the affix. If the predicate agrees with its 1 in person, then, in the normal
case, the controller and the predicate are both associated with the same person category.
IU See Keenan (1974) for an early articulation of this idea and Gazdar et al. (1985) for
following example:
(i) S- tot- on Ii Xun -e.
A3 fathrr Blsg the Xun cl
I am Xun's father.
(i) is an intransitive clause. The predicate agrees with its 1 (B 1sg suffix). But the predicate
60 CHAPTER 3
itself is a possessed nominal, and the head agrees with its possessor (A3 prefix). It happens
that both affixes attach to the same word, tot. Sentence (i) has this structure:
P
®
In terms of sponsor, the I are, E, sponsors A, and the Gen are, D, sponsors B. The Gen
arc's H arc neighbor supports B. But the 1 arc's P arc neighbor does not support A, it
R-supports it. Hence, one cannot require that the relevant support of an agreement arc be
a (direct) support, but only that it be an R-support.
What is not accounted for is the fact that a set B affix attaches to the head noun.
Nothing formulated thus far explains why it may not attach to the possessor (*S-tot Ii
Xun-on-e.). Involved here apparently is a constraint on what R-signs can be associated with
the arcs which connect the agreement arc to the relevant R-support. Apparently, they must
all be H arcs. If -on were attached to Xun, one of the connecting arcs would be a Gen arc.
12 There are some morphological facts about the agreement system which are not
formalized here. In particular, while (47) guarantees set B affixes, it does not distinguish set
B prefixes from suffixes, and hence will not guarantee that the prefixes occur on forms with
aspect prefixes, and so on.
PASSIVE CLAUSES
1. INTRODUCTION
Tzotzil has a productive passive. Its relevance to this study is that through
its interaction with other phenomena, it provides evidence for the tran-
sitivity of clauses, and allows identification of the direct object at relevant
levels.
An informal discussion of the syntax and morphology of passive clauses
is presented in section 1. Passive clauses in Tzotzil are invariably intransi-
tive, accounting for their invariable intransitive inflection. Superficial
properties of passive clauses include a set of passive suffixes, one of which
always occurs on the predicate of a passive clause, and the marking of
passive agents, when overt. Passive agents, restricted to 3rd person,
surface in two distinct ways: as objects of the preposition ta, and as
possessors of the dummy agentive noun stem -u7un. Section 2 presents a
formal account of passive clauses.
The suffix -at may be added to any transitive verb stem, yielding a passive
verb. Several facts show that clauses containing passive verbs (passive
clauses) are intransitive in the final stratum, that is, that they have no final
ergative. First, set A affixes, which must appear on the predicate of a
clause with a final ergative, cannot be affixed to passive verbs. The
following examples are active/passive pairs. The verbs in the (a) examples
are transitive and bear set A prefixes, while those in the (b) examples are
passive and cannot:
(1) a. 7i- s- maj.
cp A3 hit
He hit him.
(2) a. L- i- s- chanubtas.
cp Bl A3 teach
He taught me.
61
62 CHAPTER 4
a b c
PASSIVE CLAUSES 63
The earlier 1, a in (6), may assume the chomeur relation, in which case
it has one of two fates. It may function syntactically either as possessor of
the noun stem -u7un, or, less commonly, as object of the preposition tao
(7) Ja7 la ch- maj -at y- u7un maxtroetik ... taj yan
! cI icp hit psv A3 by teachers those other
x- chi7iltak 7une. OCK 401
A3 companions cis
Those other friends of his were beaten by the teachers.
Both -u7un and ta flag other non-term grammatical relations - the former
cause and benefactive, the latter locative and instrumental, among others
(see chapter 1, section 5).
Passive chomeurs cannot be 1st or 2nd person. Sentences in which -u7un
is possessed by a 1st/2nd person pronoun are not ungrammatical, but
ku7un and avu7un are understood as causes, not agents.
Passive clauses in which 1st and 2nd person pronouns function as objects of
the preposition ta are simply ungrammatical: 2
(12)
Cho
a b c
In the other, the earlier 1 does not occur in the second stratum:
(13)
UN b c
In both structures, the final stratum is intransitive: it contains a 1 but no 2.
The final 1 is final absolutive, and is cross-referenced on the predicate by
set B affixes.
a b c
PASSIVE CLAUSES 65
The suffixes -at and -e form passive predicates only in neutral, completive,
and incompletive aspect. In perfect aspect, the passive is formed by the
suffix -bi!:
(19) Pero vok'ol xa, vok' -bi! ta chauk. OCK 214
but cracked cl crack ppf by lightning
But it's cracked now, cracked by lightning.
the I arc local successor of a 2 arc. Hence, the passive clause (23) has the
partial structure (24).
(23) L- i- chanubtas- at y- u7un j- tot.
cp Bl teach psv A3 by A 1 father
I was taught by my father.
(24)
®
Cho C2
jtot
my father
Af St Af Af
\- chanubtas -at
teach
L L
[BI] -1-
The initial 2 arc B has a I arc local successor A which overruns C. This
configuration is the defining feature of passive clauses, and leads to the
definition in (25):
(25) Def: a is a PASSIVE CLAUSE iff a contains an arc A which is
overrun by the 1 arc local successor of a 2 arc.
Returning to (24), the successor arc A erases its predecessor arc per
the Successor Erase Law (chapter 2, law (23». (This characterizes the
68 CHAPTER 4
structure as a plain passive one (see p. 103, note 5).) As long as the initial
1 does not have a 2 arc local successor, passive structures will be finally
transitive, as desired. Vo7on 'I' heads a final 1 arc in a stratum containing
no 2 arc. Therefore, it heads a final Abs arc and sponsors a set B
agreement arc, E, in accord with chapter 3, rule (47). Arc E is supported
by its neighboring P arc, in accord with chapter 3, law (39).
The initial arc headed by jtot is a 1 arc whose persistence into the c 2
stratum would result in a violation of the Stratal Uniqueness Law. Arc C
has a Cho arc successor, a situation allowed but not required by the
Chomeur Law (see discussion in chapter 2). Hence, a grammar of Tzotzil
must assure that if C has a successor, it has a Cho arc successor. In
Tzotzil, the overrun 1 arc is either erased (if it is headed by UN), or it
sponsors a Cho arc successor. This is guaranteed by a rule like (26):
(26) Tzotzil Overrun Arc Successor Rule:
If A is overrun and has a local successor B, then B is a Cho
arc.
Rule (26) does not refer specifically to overrun 1 arcs, but to any overrun
arc, and will generalize correctly to other cases in Tzotzil, see below
chapter 7, section 7.1.
It is useful to be able to distinguish Cho arcs according to the R-signs
of their predecessors. Following RG/ APG practice, an n-Cho arc is
defined as follows:
(27) Def: A is an n-Cho arc iff A is a Cho arc and A's predecessor
is an nArc.
By (27), passive chomeurs head l-Cho arcs.
u7un NP
by
b.
L
a
ta PP
by
As these networks suggest, APG provides parallel descriptions of NP and
PP chomeurs, yielding an elegant and constrained account.
u7un
by
In accord with the Replacer Erase Law, F erases D . Structure (29)
represents the fact that chomeurs are clausal dependents in their first
stratum, but nominal dependents in the S-graph.
Arc J is a special sort of foreign successor termed PIONEER in APG.
The intuition behind pioneers is that they "create the constituent that
corresponds to their tail, .. . pioneer structures could not exist without the
pioneers themselves" (J & P, p. 605). In (29), node 100 is the tail 'created'
by its pioneer branch. Formally, neither a pioneer arc nor any of its
neighbors are initial stratum arcs. Hence, their tail node is not the head
of any initial stratum arc. J is a 'lower' pioneer, in part because it is
structurally lower than its predecessor 0 (it is a branch of a neighbour of
D). The existence of a lower pioneer entails the existence of two additional
arcs: a neighbor arc, termed a COMPANION arc, and a support arc, termed
a CLOSURE arc. Here, 1's companion is K, the arc headed by -u7un, and
1's closure is F. The closure arc is required to connect the pioneer to its
predecessor. The companion arc is motivated by the Immigrant Local
Sponsor Law (PN 12), which requires that every immigrant arc have a
local sponsor (see chapter 2, section 2.1).
K is the only possible local sponsor for J, since K is 1's only neighbor.
Since K is not an initial stratum arc, it must have a sponsor. The prede-
cessor of the pioneer arc, 0 in (29), is taken to be this sponsor, an
assumption which allows the following definition of pioneer:
(30) Def: A is a PIONEER (arc) iff there is some B such that A is a
• foreign successor of B, and if C is A's local sponsor, then B
sponsors C. (J & P, p. 605)
PASSIVE CLAUSES 71
the Flag (F) relation, borne by the preposition to the PP, and the Marquee
(Marq) relation, borne by the object of the preposition to the PP. The
structure of a PP Cho arc replacer is represented in (34) where node 100
is the PP node:
100
a
ta
by
The earlier Cho arc D has both a replacer F and a foreign successor
J. Given the Replacer Erase Law, F erases D. J is a lower pioneer: it is a
foreign successor, and its companion arc K is sponsored by its prede-
cessor, D, F is the closure arc for D and J.
A grammar of Tzotzil needs to stipulate only that a 1-Cho arc may
have a lower pioneer successor with R-sign Marq whose companion is
headed by tao Then both the existence and R-signs of the companion arc
and the support arc are determined by APG laws. The Closure Law (33)
determines the R-sign of the support. The Marq Arc Companion Law
(J & P, p. 605) requires that the companion of a Marq arc be an F arc.
All Tzotzil passive chomeurs are flagged, that is, all 1-Cho arcs have lower
pioneer successors, necessitating a rule guaranteeing this. The pioneer
arc's R-sign is either Marq or Gen - a fact which has to be stipulated only
if there are other possibilities. J & P propose that any lower pioneer arc
which is the successor of a central arc (term arc or chomeur) is either a
Marq or Gen arc. That is, the only possibilities for flagged structures are
PPs or possessed nominals (PN Law 110 (J & P, p. 622». If true, this need
PASSIVE CLAUSES 73
It remains to account for the passive suffixes, -at, -e, and -bil. To make it
possible to refer to this class of suffixes, I assume that the node represent-
ing each governs the terminal [psvj, as in:
(36)
Af
at [psv]
NOTES
I Verbs suffixed with both -be (a ditransitivizing suffix) and -at are systematic exceptions.
See chapter 7.
2 John Haviland (personal communication) suggests that the choice between ta and -u7un
has to do with definiteness (I would add humanness, as well), with ta tending to mark
chomeurs which are indefinite (or non-human), and -u7un chomeurs which are definite
and human.
3 In the Huistan dialect of Tzotzil, the agreement system distinguishes absolutives which
are Is of intransitive clauses from those which are 2s. The facts of this dialect make
PASSIVE CLAUSES 75
possible an argument that the absolutive in passive clauses is final 1. The distribution of
set B affixes in Huistec Tzotzil is similar to that in Zinacantec Tzotzil, except that where
Zinacantec Tzotzil uses a set B prefix to cross-reference transitive objects (i.e., in neutral,
completive, and incompletive aspects), Huistec Tzotzil uses both the prefix and the suffix
(the following examples are from Bricker (1977, p. 11)):
(i) Huistec Tz. Zinacantec Tz.
ch- a- j- kolta -ot ch- a- j- kolta
icp B2 Al help B2sg icp B2 Al help
I will help you.
He knows me.
which supports a pioneer is not in the initial stratum, for none of the arcs it supports are
in the initial stratum. However, nothing presently assures that a pioneer has only one
neighbor. J & P assure that a pioneer arc has no c 1 neighbor through PN Law 105, the
Pioneer Neighbor Law, which excludes any initial stratum arc as a neighbor of a pioneer.
5 As a replacer, F in (29) must have two sponsors. Its sponsors are the arcs which make it
a closure: J and D. Ar~ D, as the replaced arc, would be a sponsor in any case.
76 CHAPTER 4
I> Alternatively, one might have psv-Af arcs sponsored by some nominal arc which defines
passive clauses. However, as (25) suggests, three distinct arcs define passive clauses: the
overrun 1 arc, the overrunning arc, and the overrunner's 2 arc predecessor. Since no single
arc defines passive clauses, choosing uniquely among them is arbitrary at present. One
desirable consequence of such an approach, however, is that it might yield an explanation
of why passive affixes and other advancement morphology attach to the predicate of
the relevant clause. The explanation would involve a generalization of the condition,
formalized in the Nominal Agreement Law, ((39) in chapter 3), that agreement affixes
attach to the predicate of the relevant clause. The generalization would be that any Af arc
sponsored by a nominal arc A is an R-branch of A's P arc neighbor (excluding Af arcs
which are R-branches of the sponsoring nominal arc itself, as Af arcs headed by case
affixes would be, for example).
REFLEXIVE CLAUSES
1. INTRODUCTION
2. REFLEXIVE CLAUSES
77
78 CHAPTER 5
a
REFLEXIVE CLAUSES 79
Initial parallel 1 and 2 arcs (recall that parallel arcs share a head and
a tail) are a sufficient condition for the reflexive nominal, a condition
satisfied by the structures of (1 )-( 4).
When two initial arcs overlap, one is inevitably replaced by an anaphoric
arc whose head is an anaphoric pronoun agreeing in person and number
with the head of the relevant overlapping arcs (see section 4 for defini-
tions of italicized expressions). This is represented in (6) where C is an
anaphoric arc:
(6)
a pronoun
Arc C replaces arc B here. Replace, formally defined in chapter 2, has the
following properties. When one arc replaces another the first has the same
R-sign and tail as the second and the first coordinate of the first is greater
by 1 than the last coordinate of the second. Further, the head of a replacer
heads no initial arc.
That initial overlapping arcs always involve such replacers is not as
evident in a language like Tzotzil, where pronouns are not generally
pronounced, as it is in a language like English, but I assume that in Tzotzil
too, anaphoric pronouns are introduced as the heads of replacer arcs.
The difference between Tzotzil and English is simply that the heads of
anaphoric arcs are not generally pronounced in Tzotzil.
Of the two overlapping arcs which make an anaphoric arc possible, one
is replaced by the anaphoric arc and the other is anaphorically connected
to it. In (6), C replaces B and is anaphorically connected to A. Note that
in simple cases, where both A's head (a) and C's head (c) are pronounced,
a antecedes c. The relation anaphorically connected between arcs is
roughly analogous to the relation antecede between heads of arcs. Hence,
any arc headed by c will be anaphorically connected to a.
Given the notion anaphorically connected, a reflexive arc is an arc
which is anaphorically connected to a neighboring arc. 2 Since C in (6) is
anaphorically connected to its neighbor A, C is a reflexive arc.
80 CHAPTER 5
In many languages (e.g., German), the pronoun which heads an arc like
C in (6) surfaces as 2. However, in Tzotzil, the pronoun which heads C is
not a surface 2 (or a final 2) but genitive of the surface (and final) 2.
Hence, C is itself replaced by an arc with Gen and H branches, where -ba
heads the H branch and the anaphoric pronoun heads the Gen arc:
(7)
2 c2
Gen
pronoun;
-ba
self
Arc D is what J & P term a camouflage arc - roughly an arc headed by a
genitive/head structure which replaces an arc whose head is realized as
the genitive in question. In Tzotzil, the camouflage arcs which replace
reflexive arcs are always headed by what I term here the reflexive nominal.
Structure (7) accounts for the observed properties of reflexive clauses
as follows. They are finally transitive since the final stratum contains both
a 1 and a 2. Hence the final 1 is cross-referenced by set A affixes. The
reflexive nominal is final 2; as a possessed nominal, it is 3rd person and is
cross-referenced on the predicate by no overt set B affix. Its possessor
agrees with the 1 in person and number and is appropriately cross-
referenced on the head noun, -ba.
Finally, the meaning of reflexive clauses is neatly accounted for. All and
only those arcs which are in the initial stratum are relevant to the meaning
of the clause. Under (7), the same element is initial 1 and 2; hence, the
same element fills the semantic argument positions corresponding to those
syntactic relations. Here, it is the same group of individuals who do the
'shutting in' and are 'shut in'.
I assume the following rule, stated informally.
REFLEXIVE CLAUSES 81
3. RECIPROCAL COREFERENCE
(12)
2 c2
pronoun;
Such replacement is routine in the case of overlapping initial arcs, and is
the source of coreferential pronouns (see chapter 2, section 2). In (12),
B is replaced by a pronominal arc, C, sponsored by both A and B. Arc C
replaces B, and therefore erases B per the Replacer Erase Law (chapter 2,
(22)). Arc A is its seconder.
Furthermore, C is an ANAPHORIC arc:
(13) Def: A is an ANAPHORIC arc iff A is a pronominal arc and A
has two sponsors.
Anaphoric arcs are thus distinguished from ghost arcs which are headed
by dummy pronouns and have a single sponsor. An anaphoric arc is
ANAPHORICALL Y CONNECTED to its seconder, as are all its successors.
(16)
@
2 c,
@
1c u 2 c)
pronouni
sbaik
selves
Here, the reflexive arc C has a Gen arc lower pioneer successor F. C is
replaced by F's support, D, which is a CAMOUFLAGE arc:
I focus now on the relation between the Gen arc branches of camouflage
arcs, and their precedessors. Informally, the pronoun which replaces the 2
surfaces as possessor of the construction-specific noun -ba. The situation
is exactly parallel to that of passive chomeurs, which are realized as
possessors of the construction-specific noun -u7un. Compare their relevant
sub-PNs in (18):
84 CHAPTER 5
(18) a.
®
H@
-ba
self
b.
-u7un
by
In both cases, an arc has as successor a branch of its own replacer. Arc
A's successor is B, while its replacer is C; D's successor is E, while its
replacer is F. Both Band E are lower pioneers. A lower pioneer is always
paired with a companion arc. In the case of reflexives, the H arc headed
by -ba is that companion arc.
Most of the structure represented in (18a) follows from APG laws
governing lower pioneer structures. The existence of the replacer arc C
and the companion arc G follows from the Closure Law as does C's
REFLEXIVE CLAUSES 85
R-sign (see p. 71). As noted earlier, G's R-sign (H) may follow from
general restrictions on the distribution of Gen arcs.
NOTES
I The noun -ba exists independently in Tzotzil and means 'top, face'.
The intuition behind the term reflexive arc is that in particular languages, so-called
reflexive pronouns generally head some subset of reflexive arcs, as here defined.
] Further, it assumes that all cases of parallel initial 1 and 2 arcs involve a reflexive arc.
Otherwise, coreference between the initial 1 and 2 would be possible in a clause not
containing the reflexive nominal. However, this is not possible. See note 7.
86 CHAPTER 5
4 There is another reflexive clause type with reciprocal semantics which this account may
(i) A is a PRO(NOMINAL) arc iff A is a graft and A is a nominal arc and there is
no B which sponsors A and is supported by A.
A GRAFT is an arc with no overlapping sponsor. Hence, a graft is not an initial arc, nor is
it the successor of any arc. The last condition distinguishes pronominal arcs from closure
arcs, which would otherwise satisfy the definition. Closure arcs are sponsored by one of
their branches in 1 & P.
6 Under (17), those Tzotzil passive chomeurs which function as possessors of -u7un
define camouflage structures. The closure arc in such structures is a camouflage arc.
7 The statement of (19) assumes that in all cases of parallel initial arcs, one of the two arcs
will be replaced by a pronominal arc (which will, by definition, be a reflexive arc), and that
it is only necessary to specify which reflexive arcs are replaced by camouflage arcs. The
implication then is that all other cases of coreference between clausemates involve
anaphoric pronouns not of the form -ba, i.e., zero pronouns. In at least one case, discussed
in chapter 7, coreference between clausemates is impossible, requiring an additional
condition.
UNACCUSATIVE CLAUSES
1. INTRODUCTION
Within RG, it has been proposed that there are two classes of intransitive
predicates: those which take initial 1s ("unergative predicates") and those
which take initial 2s ("un accusative predicates") (Perlmutter 1978). In
unaccusative structures, the 2 generally advances to 1 satisfying the Final
1 Law (this advancement termed "unaccusative advancement"):
(1)
(6)
©
2 C2
pronoun;
90 CHAPTER 6
®
P CU .3
Gen
pronoun;
Ii 7ixime; 7iskap
corn mix -ba
The reflexive construction is found only when the same nominal happens
to be 1 and 2. However, verbs which occur in unaccusative reflexive
clauses often do not occur in transitive non-reflexive clauses. The verb je7
only occurs in reflexive clauses:
(9) a. Ta x- je7 s- ba.
icp ramify (?) A3 self
It's putting forth branches.
b. *Ta s- je7.
icp A3 ramify ( ?)
Je7 apparently occurs only as predicate of an initially unaccusative clause.
Only those unaccusative verbs which happen also to occur in initially
transitive clauses (in transformational terms, have dual subcategorization)
will occur in both unaccusative reflexive clauses and non-reflexive transi-
tive clauses. Verbs which occur only in unaccusative reflexive clauses
include: 2
(10) je7 -ba 'put forth branches and leaves'
k'uyan -ba 'form clumps' (people in crowd)
lajtzan -ba 'break out' (measles)
likan -ba 'hang heavily' (rain cloud)
lin -ba 'fill out, mature physically' (child, tree)
vajan -ba 'erupt' (pox, measles)
tzop -ba 'become puffy' (face), 'swarm' (caterpillars)
tzot' -ba 'shrivel' (fruit, skin)
tz'ay -ba 'become pretty and neat'
All these verbs must be restricted to clauses whose initial stratum is
unaccusative. In transformational terms, they subcategorize a 2 and no 1.
In contrast, ni7 and nijp'un (see (2)-(3)) occur both in initially transitive
and initially unaccusative clauses. Verbs which occur with two different
sets of initial nuclear terms will be termed BIVALENT STEMS. Further,
those like ni7 and nijp'un which are reflexive when unaccusative are
termed REFLEXIVE UNACCUSATIVE STEMS.
Like passive verbs, the verbs of these clauses are intransitive (note the
lack of any Set A (ergative) prefix), and are systematically related to
transitive stems:
Finall)" the 1s of these clauses, like those of passive clauses, have the same
thematic relation as the 2s of the corresponding transitive clauses.
These verbs are unlike passives in several respects as well. For one
UNACCUSATIVE CLAUSES 93
thing, they are identical to the associated transitive verb. For another, they
cannot cooccur with an agent phrase:
Haviland (1981, p. 253) notes that verbs of this type can take instrument
complements introduced by ta, or a complement understood as 'cause'
whose presence is registered by the verbal clitic 70 (see pp. 8-9):
Haviland says about (23): "By it was closed by Xun we do not mean that
Xun actively closed the house, but that somehow the house closed on
account of him: if, for example, he fell against the door, or if this body was
blocking it." [translation mine]
It is possible that these verbs have a syntactic argument interpreted as
agent which, for some reason, cannot occur in surface structure. There is
some support, however, for the view that they have no such syntactic
argument. Although (11)-(16) seem to implicate an agent, this has,
arguably, nothing to do with grammar, but rather with the situations
described: purchases, abduction, and construction all require agents.
Further, some of these sentences, namely (14) and (16), have a non-
agentive interpretation, as in English. In contrast, the clauses in which
passive verbs occur always entail the existence of an agent. Haviland
(1981, p. 258) contrasts the meanings of the following pair, where the first
is the passive of a perfect, and the second the perfect of an intransitive
stem of the type-under discussion: 3
(24) a. Mak -bi! Ii na -e.
close ppf the house cl
94 CHAPTER 6
b. 7i- ve7.
cp eat;\,
He ate/*It was eaten.
The mixed final transitivity of plain bivalent stems has two morphological
consequences.
3.2.1. Perfect
Since plain bivalent stems occur both in finally tranSitIVe and finally
intransitive structures, they form perfects both with -em and -oj, with
associated differences in meaning. They also form a perfect passive since
they can be predicates in passive clauses:
UNACCUSATIVE CLAUSES 97
b. muk 'bury'
muk-em 'it's buried'
s-muk-oj 'he's buried it'
muk-bil 'it's been buried'
d. jam 'open'
jam-em 'it's open'
s-jam-oj 'he's opened it'
jam-bil 'it's been opened'
Note that ve7 also forms perfects with all three suffixes, but the semantic
relations among the three are different:
(31) ve7 'eat'
ve7-em 'he's eaten'
s-ve7-oj 'he's eaten it'
ve7-bil 'it's been eaten'
Suffixation of -em to a transitive stem which has a shape other than
eve results in an ill-formed word, for these stems cannot be finally
intransitive:
(32) *7elk'an-em < 7elk'an 'steal' (tv)
*meltzan-em < meltzan 'make, construct' (tv)
*k'opon-em < k'opon 'address' (tv)
*70jtikin-em < 70jtikin 'know' (tv)
*kolta-em < kolta 'help' (tv)
3.2.2. Subjunctive
b. Laj s- pas.
end A3 make
He finished making it.
See also chapter 11, section 2.4.
4. VERB CLASSIFICATION
TABLE V
Verb Classes
The other possibility is that the 2 arc be erased by its 1 arc successor
(see chapter 2, pp. 26-30). This is allowed by the Successor Erase Law
(chapter 2, (23)), which requires that a successor erase its predecessor
unless that predecessor has a replacer (in which case, it is erased by the
replacer). The PN in (37) illustrates this:
100 CHAPTER 6
(37)
®
1 c2
(39)
® ©
pronoun; -ba
The key difference then between a structure like (39) and a core-
ferential one like that represented by (16) in chapter 5 is that in (39), the
UNACCUSATIVE CLAUSES 101
NOTES
I This list is based on Laughlin (1975), and by no means includes all relevant examples.
Laughlin recognizes a category "reflexive verb" to which the second verb in each pair
below is assigned.
UNACCUSATIVE CLAUSES 103
tive stems by infixing [hI before the final consonant. In the sister language Tzeltal, [hI shows
up before voiceless stops and affricates, deleting elsewhere. E.g., buht 'it filled up'/but 'to
fill it up', pas < pahs 'it is done'!pas 'to do it' (Kaufman 1971). In Tzotzil, preconsonantal
[hI has been lost, resulting in homonymy between monosyllabic transitive stems and their
corresponding unaccusatives.
5 It should be clear now why it is necessary to specify that Tzotzil passives, which also
involve parallel 1 and 2 arcs, are plain and not reflexive. Rule (i) will do this, given the
definition of passive clause in chapter 4:
(i) Tzotzil Plain Passive Rule:
If A is the 1 arc local successor in a passive clause, then A erases its
predecessor.
APPENDIX TO CHAPTER 6
Plain Unaccusatives:
(1) Ja7 la tz- na7 -ik ti ch- poj. OCK 193
cI icplA3 know pi the icp take away,v
They knew that she had been abducted.
1. INTRODUCTION
2. DITRANSITIVE CLAUSES
A Tzotzil transitive clause can contain a 3 (indirect object) just in case the
predicate is suffixed with -be. Without -be, transitive predicates take only a
1 and 2, as shown with 7ak' 'give' in (1):
Adding -be to the verb allows for a third un flagged nominal argument:
(2) 7a Ii Xun -e, ba y- ak' -be chitom Ii 7antz -e.
topic the Xun el go A3 give io pig the woman el
Xun went to give the pig to the woman.
Further, a verb suffixed with -be requires a third argument. Even without
any overt arguments, 7iyak'be means 'x gave y to z' where x, y, and z all
refer to specific individuals.
The following examples are like (1 )-(2): the verbs of the (a) sentences
are not suffixed with -be; these clauses contain only a 1 and a 2. The verbs
of the (b) sentences are suffixed with -be; the clauses contain three
nominal arguments. The (c) sentences show that without -be, a transitive
verb can take only two arguments.
(3) a. 7i- j- meltzan J- p'ej na.
cp A I make one nc house
I made a house.
b. Tz- paj -be taj 7akuxa taj ka7 7une. OCK 382
icp/A3 push io that needle that horse els
He's pricking the horse with the needle.
(i.e., He's pushing the needle into the horse.)
3. 3-TO-2 ADVANCEMENT
3.1. Agreement
(7) Meltzan -b- [0] -on lek i garafon -e. OCK 203
fix io imp BIsg good the jug cl
Fix the jugs carefully for me.
In (8), -a- 'B2' cross-references the thematic recipient:
(8) Ch- a- k- ak' -be. OCK 75
icp B2 A J give io
I'll give it to you.
In (9), -ik 'pI' cross-references the thematic benefactive:
(9) Ch- a- j- mil -be -ik. OCK 131
icp B2 Al kill io pI
I'll kill them for you (pI).
The agreement facts of (7)-(9) follow from (6) and the agreement rules of
chapter 3. The initial 3 advances to 2 and, as final absolutive, is cross-
referenced by set B affixes and optionally by -ik.
When the verb of a ditransitive clause bears no overt set B affix, it must
have a 3rd person final absolutive. Such clauses are understood to have
3rd person recipient, benefactive, or target, etc. Again, this follows from
(6) and the agreement rules of chapter 3.
p 2
Cho
p
Structure (15) wrongly claims that ditransitive passives are finally transi-
tive. The analysis in (14), on the other hand, accounts neatly for the final
intransitivity of ditransitive passives by including 32A, thus entailing final
chomeurhood for the initial 2.
Further, structure (14) predicts, correctly, that the initial 2 cannot
control agreement in ditransitive passives. It cannot be cross-referenced by
set B affixes: ch-i-7ak'-b-at (icp-BI-give-io-psv) (as in (13)) cannot mean 'I
was given to him.' And the plural suffix -ik cannot cross-reference the
initial 2:
(16) 7i- 7ak' -b -at -ik.
cp give io psv pi
They were given it/them/her/him.
In (16), the recipient must be interpreted as plural, while the number of
the initial 2 is not restricted.
(14) also accounts properly for the word order in ditransitive passives.
The following three principles are motivated exclusively by word order in
active clauses:
(17) P(redicate)s precede non-Ps
Is follow 2s
2s follow unflagged chomeurs
These principles entail the following schema for word order in active
ditransitive clauses:
(18) P - unflagged Cho - 2 - 1
(Three full nominals rarely occur overtly after the predicate, but (18)
110 CHAPTER 7
correctly predicts the relative order of any two. The chomeur is specified
as unflagged because the position of flagged chomeurs (as in passives) is
not fixed.) Under (14), (18) also correctly predicts the word order in
passive ditransitive clauses:
(19) P - unflagged Cho - 1
Sentence (20), with the structure In (21), exemplifies word order In
ditransitive passives.
(20) 7ak' -b -at Jun syen soltaro Ii J- chamu7
give io psv one hundred soldier the agn chamulan
preserente 7une. OCK 103
president cls
The Chamulan president was given one hundred soldiers.
(21 )
Ditransitive reflexive clauses exist and are characterized both by the suffix
-be, and a form of the reflexive nominal. In such clauses, most commonly
associated with the semantics of reciprocals, the initial 1 and 3 are
coreferential.
(22) 7i- j- mal -be j- ba Ii kalto -e.
cp Al spill io Al self the broth cl
I spilled the broth on myself.
DITRANSITIVE CLAUSES 111
kalto I I 7ijmalbe
broth spill
Cho C 3,4
ijmalbe
spill
kalto I I
broth
jba
self
7iyak'be Ii mayoletike
gave police
sba
self
5. RESTRICTIONS ON ADVANCEMENT
fact that no intransitive verb stem can be suffixed with -be shows this.
Intransitive verbs like k'opoj 'address' or manolaj 'sell' might be expected
to govern initial 3s, which would necessarily advance to 2, yielding a final
transitive stratum:
(34)
the women, or They will sell you to the owner cannot be translated by
ditransitive clauses:
(50)
7ak'bilon kot I UN
was given my tortillas
Since -bil forms the passive perfect of both transitive and ditransitive
stems, sentences like (51) are ambiguous:
(51) 7ak' -bil -on.
give ppf Blsg
I was given Ito someone].!1 was given it.
On the monotransitive reading, 'I' is initial 2 and final 1 (by passive). On
the ditransitive reading, T is an initial 3 which advances first to 2, and
then to 1.
In APG terms, a clause involving 32A contains a 3 arc with a 2 arc local
DITRANSITIVE CLAUSES 119
requires that the successor of any locally overrun arc be a Cho arc, and
hence applies both to passive and to 32A.
Finally, chomeurs in clauses involving 32A are 3rd person, as are
passive chomeurs. The following generalization is possible.
(55) Tzotzil Cho Arc Rule:
If A is a Cho arc, then A's head is 3rd person.
These rules account for the properties of ditransitive clauses, and
they interact properly with rules governing passive and reflexive· clauses
to account for the properties of ditransitive passives and ditransitive
reflexives.
(56)
100
CD
®
jun tzeb
a girl
UN
Cho C'.4
7ijmalbe
spill
Ii kaltoe
broth
I
jba
self
122 CHAPTER 7
The key feature here is the overlap of the initial 3 arc B with the initial
arc A, yielding the co referential reading. As overlapping initial arcs, A
and B sponsor a pronominal replacer C, also a reflexive arc. As required
by the Coreferential Arc Law (see chapter 2, page 30), C replaces a c,
arc. Arc C is a free 3 arc (it has no replacer) and has, per rule (52), a 2
arc local successor (D) which overruns a 2 arc. Given the conditions on
camouflage reflexive arcs (chapter 5, rule (19», the 2 arc successor D
must be replaced by the camouflage arc, G. For D is a 2 arc, anaphorically
connected to a 1 arc (A). Arc D is anaphorically connected to A because
D is the successor of an arc seconded by A, namely C.
A restriction limiting the reflexive nominal to the heads of 2 arcs was
motivated in section 3.3 by several sets of facts. In present terms, this
involves a restriction on those arcs which are headed by the reflexive
nominal, namely, camouflage arcs which replace reflexive arcs. (Note that
not all camouflage arcs are so restricted, for certain passive chomeurs
head camouflage arcs.) Also constrained are the R-successors of camou-
flage arcs which replace reflexive arcs. This accounts for the fact that the
reflexive nominal cannot be extracted, i.e., head an overlay arc. The term
REALIZATION will be useful here. It picks out those arcs which are
connected to an arc by a chain of relations whose links may be the
successor or replacer relation:
(58) Def: A is a REALIZATION of B iff B R-sponsors A.
7iyak'be
gave
Ii mayoletike
police
sba
self
DITRANSITIVE CLAUSES 123
Arc C is the anaphoric arc which replaces the initial 2 arc B, and It IS
a reflexive arc since it is anaphorically connected to a neighboring 1 arc
(A seconds C). In accord with (chapter 5, (19)), C is replaced by a
camouflage arc, D. Since C sponsors D, D is a camouflage realization of a
reflexive arc, and it satisfies (59). Arc C, however, has a Cho arc successor
E which is also a camouflage arc (since it supports a Gen arc lower
pioneer), and a realization of a reflexive arc (since C R-sponsors it).
Hence (60) violates (59), thus accounting for the fact that (29) does not
have a reading on which the initial 1 and 2 are coreferential.
8. CONCLUSION
NOTES
I A more limited constraint would require that any final arc a reflexive nominal heads be
a 2 arc. Rule (31) is preferable for two reasons. First, as will become clear in chapter 8,
section 2.3, the reflexive cannot head a genitive arc either. Rule (31) obviates the need for
an additional constraint, while a version of (31) restricted to the final stratum would not.
Second, the proposed revision fails to account for the overlay facts noted below in the text,
for the reflexive heads final 2 arcs in all those examples.
2 Postal (1985) argues that such structures exist, e.g., in English, terming them tertiary
passives.
) Cognates of -be function productively in a number of Mayan languages to mark
advancements to 2. In some (e.g., Tzotzil, Tzeltal, Chol) advancement is from 3, but in
others (e.g., Quiche), it is from Instrument. A number of languages which historically
marked advancement of Instrument to 2 with cognates of -be have apparently lost the
advancement rule, but the affix persists, registering the presence of an instrument. See
Norman (1978).
4 Ditransitive perfect passives are potentially problematic for this rule. Recall that they are
formed with the suffix -bi!, the same suffix which forms monotransitive perfect passives.
Thus: as noted earlier, 7ak'bilon is ambiguous, meaning either 'I was given' or 'I was given
it'. On the first reading, the clause is monotransitive and involves only passive. In this case,
124 CHAPTER 7
-bit is psv and satisfies the condition that every passive clause contain a psv suffix. On the
second reading, the clause is ditransitive and involves both 32A and passive. The question
is whether in such cases -bit is only psv, in violation of (53), or is both psv and io. In the
interests of satisfying (53), I assume -bit must be psv and may be io. Rule (53) will rule out
ditransitive passives in which -bil is not io, as well as monotransitive passives in which it is.
APPENDIX
Ditransitive Actives:
(12) Poj -b -at -ik 7un y- u7un taj pukuj 7une. OCK 355
remove io psv pi cI A3 by that devil cis
It was taken from them by that devil.
1. INTRODUCTION
2. POSSESSOR ASCENSION
(4)
chistoyilanbe he
keep lifting
jjol I
head
Crucial properties of (4) are that the initial stratum contains no 3, but
does contain a 2 which has a genitive. This 2 is the 'ascension host'. The
genitive ascends to clausal 3 in the second stratum, and as a 3 necessarily
advances to 2, putting the initial 2 in chomage. Given what has already
been established about Tzotzil grammar, most of the properties of (1 )-(3)
are deducible, under the assumption that their structures are properly
represented by (4).
First, (4) predicts correctly that the predicate will be suffixed with -be,
since the clause contains a 3 (and therefore involves 32A). Second, (4)
predicts that such clauses will always contain an overt nominal which is
possessed, this nominal being the host of the ascension. J-jol, k-ol, and
I-a-tot-e are the nominals in question in (1)-(3). They must be overt
because only pronouns may fail to surface, and a possessed nominal
cannot be a pronoun. Third, because these nominals are possessed, each
will bear a set A agreement affix marking agreement with its possessor.
This is true of j-jol, k-ol, and I-a-tot-e. (The possessor itself will generally
not appear overtly if it is pronominal, as in these examples.) Fourth, since
the 2-possessor is itself the final 2, it will control set B (absolutive)
agreement. In (1), for example, the 2-possessor is the 1st person pronoun,
and that nominal is also final 2. Hence, the verb agrees with it by the 'B 1'
affix. A recurrent feature of (1 )-(3) is that the set B affix on the predicate
matches in person the set A affix on the possessed noun. Thus in (1), both
ch-i-s-toyilan-be and j-jol have 1st person agreement affixes. The affix i is
128 CHAPTER 8
Bl and j is AI. This is not accidental, since in (4), the initial 2-possessor
is final clausal 2, and thus agreement is controlled in the two domains by
the same nominal.
The fact that (4) contains no initial 3 amounts to the claim that the
nominal which is 3 in this construction makes no semantic contribution
beyond that made in its function as 2-possessor. This is not to say that
there are no cases in which one could not point to some independent
thematic relation which the 3 might bear (e.g., malefactive in (2)), but that
there are cases where it bears none. The most convincing cases are those
in which the possessor refers to an inanimate nominal, one unlikely to be a
benefactive, malefactive, etc.
As in many other languages, syntactic possessors are understood in a
variety of notional relations to the possessed noun. Among these are
ownership (my cat), body-part (my arm), kin (my son), and others. There
is also a peculiarly Mayan category termed inanimate possession,
described by Laughlin (1975, p. 25) as indicating "the linkage of two
objects or of an object and an action either by location or by design." This
category is formally distinguished from other categories of possession by
the presence of a suffix which is otherwise absent on the possessed noun.!
For example, the noun tzek 'scorpion' can be possessed in the usual
manner, and the result, I-a-tzek means 'your scorpion', i.e., your pet. If the
suffix -al is also added to the noun, the result is interpreted semantically
as so-called 'inanimate possession': I-a-tzek-al 'your scorpion', i.e., the
scorpion that almost bit you (see example (7a) below). Similarly, k-alak'
'my chicken', i.e., the one I own, contrasts with k-alak'-il 'my chicken', i.e.,
the one which will be used for my curing ceremony, or the chicken that
pecked me, etc. (The suffix which appears on the head noun when
inanimately possessed has the form -VI; V = vowel, and is glossed 'poss'
in examples.) It is not clear how to characterize inanimate possession
semantically, but the relation between the noun and its possessor seems
more abstract than in other cases of possession. Despite the name,
inanimate possession does not always involve inanimate possessors (cf. the
examples just discussed).
The meaning of inanimate possession is seen most clearly in clauses
containing no 3. The following (from Haviland 1981, pp. 196-8) are all
intransitive; the 1 is inanimately possessed. Haviland distinguishes several
types of inanimate possession. The possessor may be related to the head
noun by location:
2.1. Agreement
nominal of the same person and number as the initial 2-possessor, as (4)
predicts. Of course, in clauses not involving PA, this is not generally the
case:
(11) *Ch- i- s- toyilan j- jol.
icp BI A3 keep lifting Al head
(He kept lifting my head.)
2.2. Passive
There are passive versions of clauses involving PA. In these, it is the final
1 that matches the possessor of the initial 2 in person:
chitoyilanbat UN
kept being lifted
jjol I
head
The initial 2-possessor raises to 3, advances to 2, putting the initial 2
(the host) in chomage, and then advances to 1, in this case keeping the
unspecified lout of c4 • The verbal suffixes -be and -at mark both
advancements. Like other ditransitive passive clauses, (15) is finally
intransitive (the 1 is cross-referenced by set B affixes), and contains two
final unflagged arguments: the final 1, and the final chomeur.
2.3. Reflexives
Given what has already been estahlished about ditransitive reflexives
(chapter 7, section 3.3), we may ask whether we should expect to find
examples of PA in ditransitive reflexive clauses. Ditransitive reflexives
express coreference between 1 and 3, so if there are examples involving
POSSESSOR ASCENSION 133
(20)
70 7iyik'be
H marry
sba
selves
134 CHAPTER 8
(30)
(Sentence (31) has the irrelevant reading 'Did your father sell you pigs?',
where the 2nd person pronoun is initial 3 and thematic recipient, benefac-
tive, or malefactive.) I conclude then that the possessors of 1s cannot raise
to 3.4
Further, the P A host must not only be a 2, but a 2 in a transitive
stratum, that is, it must occur with a 1. Such 2s are ACCUSATIVES. The
distinction is necessary because 2s in unaccusative clauses do not host P A.
To see this, consider the Tzotzil analogue of Her husband was buried,
with the structure in (33):
(33)
2 p
7imuk
was buried
smalal her
husband
(35)
7imukbe
was buried
smalal her
husband
138 CHAPTER 8
At this point, we can establish necessary conditions for PA. There are
three: that the possessor ascend as a 3, that its host be an accusative, and
that it not be reflex-anteceded by the 1 of the clause it ascends into.
Informally:
(40)
1
1
he H Gen
I liyak'be
gave
skaltoal it
broth
The result of raising the possessor of skaltoal to 3 is a structure in which
the initial 3 is final chomeur. This is true whether the initial 3 is itself put
in chomage (as in (41a) below), or whether the initial 3 advances to 2, and
is put in chomage by the advancement to 2 of the raised 3 (as in (41 b)):
(41) a.
he I
skaItoal it
broth
140 CHAPTER 8
(41) b.
he
skaltoal it
broth
Since the 1st person pronoun is neither final ergative nor final absolutive,
it is not cross-referenced on the verb. But neither structure in (41) is
realized - both are ill-formed. Example (42) lacks the intended reading,
whether vo7on occurs in surface structure or not: 5
(42) *7a Ii Xun -e, 7i- y- ak' -be s- kalto -al
topic the Xun cl cp A3 give io A3 broth poss
(Ii v070n -e).
the I cl
(X un gave me its broth.)
Consider also the Tzotzil for 'Petul gave my sheep to Xun':
(43) 7a Ii Petul -e, 7i- y- ak' -be j- chij Ii Xun -e.
topic the Petul cl cp A3 give io AI sheep the Xun cl
Petul gave my sheep to Xun.
Here, Ii Xune is initial 3 and final 2. If the 2-possessor ascended to 3 and
advanced to 2, it would be final absolutive and would control absolutive
agreement:
(44) *7a Ii Petule, 1- i- y- ak' -be j-chij Ii Xun-e.
cp BI A3 give io
But (44), which has one of the structures in (41) (with different terminals,
of course), is ill-formed, and has no reading.
In part, the problem with the structures in (41) seems to be that they
contain two 3s; one in the first stratum and one in the second. However, a
POSSESSOR ASCENSION 141
requirement that no clause contain more than one 3 arc is too strong,
because the analysis of ditransitive reflexives involves clauses with two 3
arcs; the initial 3 arc (parallel to the 1 arc) and the anaphoric 3 arc which
replaces it. What distinguishes the case of ditransitive reflexives from that
involving PA is that in the former one 3 arc replaces the other, while in
PA cases, the two arcs are not thus linked. Recalling that a free arc is one
which is not replaced, the restriction is that a clause contains at most one
free 3 arc.
Rule (45) also accounts for the fact that no Tzotzil clause has two initial
3s. Such clauses will always contain two free 3 arcs. A structure containing
two initial 3s, one interpreted as recipient, the other as benefactive (e.g.,
the structure of 'He gave the pigs to Xun for Maruch'), has no realization:
(46) a. *7i- y- ak' -be chitom Xun Ii Maruch -e.
cp A3 give io pig Xun the Maruch c/
impossible under reflexive coreference. This is stated in (38) and holds for
all structures involving PA, regardless of the person of the possessor and
whether the possessor is a pronoun or not. The fact that the sentences in
(52) lack the noncoreferential reading is a more restricted phenomenon.
Not all monotransitive clauses containing possessed 2s entail reflexive
coreference between the 1 and 2-possessor, as (47)-(49) and (50) show.
This holds only when the possessor is a 3rd person pronoun, and even
then there are exceptions.6 Hence, while (51) and (52) are minimal pairs,
forced coreference and forced noncoreference are not complementary.
Noncoreference is forced whenever PA is involved, while (reflexive)
coreference is forced only when PA is not involved and the possessor is a
3rd person pronoun. For that reason, distinct rules account for forced
co reference and forced noncoreference.
There are two ways to think about sentences like those in (52). From
one point of view, these sentences have only coreferential interpretations
because PA is forced otherwise. From the other, (52a,b) are unambiguous
because under certain conditions the 1 must reflex-antecede the 2-pos-
sessor, and (52a,b) satisfy those conditions. The first approach suggests
a rule forcing PA under conditions of noncoreference, leaving the co-
referential cases. But this entails two problems, in addition to that raised
by footnote 6. There are in fact two situations in which PA is not forced
even where the 1 and 2-possessor are not coreferential. One cannot be
discussed until chapter 9, section 4.1. The other is when the clause already
contains a 3, and ascension of the possessor as 3 would yield a structure
containing two free 3s, in violation of (45). The following sentences
illustrate this:
(63) /
/
/
©
3 c2
H Gen~========~
Recall that foreign successors are immigrant arcs, and that immigrant arcs
have two sponsors: a foreign one and a local one. The predecessor arc
itself is the foreign sponsor (B in (63)). A's local sponsor is C, as
represented in (63). We will see below that this is lawfully determined, and
requires no stipulation. It is desirable that C and A be formally connected,
as they are through the sponsor relation, because C determines certain
properties of PA generally, and certain properties of Tzotzil PA in
particular.
What in (63) needs to be stipulated for Tzotzil, and what is determined
by general principles? At least some of the R-signs in (63) must be
determined by ljlnguage-specific rules. Both the fact that possessors
ascend to 3 (rather than to 2, say) and the fact that they ascend only out
of 2s appear to be language-specific. (In Southern Tiwa, for example, pos-
sessors ascend out of 2s, as 2s, and they ascend out of intransitive Is as
well (Allen, Gardiner, and Frantz 1984)l On the other hand, the fact that
148 CHAPTER 8
the possessor raises into the clause containing the host is not language-
specific.
An APG law requires that an ascendee raise into some clause which
contains the host constituent, but not necessarily into the clause which
(immediately) governs it (the Nominal Arc Immigrant Local Sponsor Law,
J & P, p. 706). This laxity is permitted because there are ascensions which
are not restricted to adjacent constituents (e.g., English Object Raising).
However, in Tzotzil PA, and in PA constructions generally, the possessor
ascends into the clause which (immediately) governs the host. Therefore I
propose the following law which refers specifically to PA, and determines
that a raised possessor will ascend into the clause which governs the
possessed nominal. It does so by limiting the local sponsor in PA to the
arc headed by the possessed nominal. The local sponsor in an ascension
determines what clause the ascendee is raised to (by definition).
(64) Possessor Ascension Local Sponsor Law:
If A is a Gen arc with a Term arc foreign successor B, then A
is a branch of B's local sponsor.
The Tzotzil PA rule must guarantee one more thing in addition to the
fact that the possessor raises as a 3 and raises only out of an (accusative)
2. Namely, that the possessor and 1 of the governing clause are not
reflexive coreferents. As emphasized in earlier discussion, reflexive and
reciprocal coreference have, by assumption (see p. 81), the same syn-
tactic representation in Tzotzil: both involve overlapping initial stratum
arcs. One of the two overlapping arcs is replaced by an anaphoric are,
the other seconds it. The anaphoric arc is anaphorically connected to
its seconder, as are all its successors. To distinguish reflexive and reci-
procal coreference, I will say that one case involves REFLEX ANAPHORIC
CONNECTION, the other RECIP ANAPHORIC CONNECTION (or more con-
veniently, REFLEX-CONNECTION and RECIP-CONNECTION). (Of course,
this simply allows reference to the difference between the two, and in no
way offers any insight into the difference.) In clauses which involve PA,
the immigrant 3 arc cannot be reflex-connected to its neighboring 1 arc.
Rule (65) properly constrains Tzotzil PA:
(65) Tzotzil Possessor Ascension Rule:
If A is the Term arc foreign successor of a Gen arc, the A is a
3 arc and A's local sponsor is an accusative arc, and A is not
reflex-connected to its neighboring 1 arc.
Clauses involving PA are subject to (65), and also, since they contain 3
arcs, to all constraints on ditransitive clauses. As in other ditransitive
clauses, the 3 must advance to 2, to satisfy rule (52) of chapter 7 which
requires that any free 3 arc have a 2 arc local successor. Consider then
the representation of (1) (Chistoyilanbe iiol 'He lifted my head'):
POSSESSOR ASCENSION 149
(66)
he chistoyilanbe
keep lifting
jjol I
head
The Gen arc B has a term arc foreign successor (A), characterizing (66) as
a PA structure. In accord with (65), arc A is a 3 arc, and A's local sponsor
C is an accusative arc. The conditions on ditransitive clauses are also
satisfied. The 3 advances to 2, putting a 2 in chomage. Formally, A has a 2
arc successor E; E overruns C and C has a Cho arc successor. The final
stratum is c3, and as 2 in that stratum, the 1st person pronoun controls set
B agreement on the predicate, satisfying the agreement rules of chapter 3.
The final 1 and 2 in the clause are pronouns and are not pronounced. The
arcs they head, F and E, self-erase. Only G and I, headed by chistoyilanbe
and iiol, respectively, are surface (unerased) arcs. In (1), the predicate
occurs clause-initially, followed by the final chomeur.
In corresponding passive constructions, the analogue of E in (66)
would have a 1 arc successor, which would overrun a 1 arc.
In a ditransitive reflexive involving PA, conditions on PA, ditransitive
clauses, and reflexive clauses must all be satisfied. As illustration, consider
(19), repeated below, and its structure in (67):
(19) 7i- y- ik' -be s- ba y- ixlel -ik.
cp A3 marry io A3 self A3 y.sister pi
They married each other's younger sisters.
150 CHAPTER 8
(67)
7iyik'be
marry
Gen
@
younger
sisters H
sba
each other
The meaning of (19) is associated with the initial arcs in (67): A, J, and K,
and K's branches Band L. A key fact here is that the initial Gen arc B in
the nominal constituent 70 overlaps with the clausal 1 arc A, yielding the
coreferential reading. Arcs A and B cosponsor arc C, accounting for the
presence of the anaphoric pronoun in (19). This pronoun, which is
anteceded by the 1, controls agreement on yixlelik. Further, it is this
pronoun which ascends to 3, advances to 2, is replaced by the camouflage
nominal, and ultimately surfaces as genitive in that nominal. Formally, C
has a 3 arc foreign successor, D (in accord with (65», and as chapter 7,
rule (52) requires, D has a 2 arc local successor, E. All of the arcs C,
D, and E are anaphorically connected to A (recip-connected). Since E is a
2 arc anaphorically connected to its neighboring 1 arc, chapter 5, rule (19)
requires that it must be replaced by a camouflage arc, arc F.
Extracting the final stratum (c 4) arcs in (67) yields (68):
POSSESSOR ASCENSION 151
(68)
11. CONCLUSION
NOTES
because subsequent 32A would not satisfy the requirement that 32A create a chomeur, per
(41 ).
5 Example (42) is ungrammatical ifvo7on is overt. Without vo7on, (42) is grammatical, but
means 'Xun gave him its broth' - a sentence which contains no 1st person pronoun at all.
6 Example (i) is a textual example in which there is no PA, the possessor is a 3rd person
pronoun, and still there is no coreference between the 1 and 2-possessor (the only clear
textual example I have).
(i) It [the magic sashl did it. It swept the inside of the house. It tidied up the inside
of the house. It looked over the inside of the house. It took the dishes.
S- suk' la ti s- pulatu s- boch x- cho7 y- ek'en. OCK 207
A3 rinse cl the A3 plate A3 bowl A3 metate A3 metate platform
It washed her plates, her bowls, her metate, her metate platform.
POSSESSOR ASCENSION 153
It is unclear what factors here override the co reference condition, but they appear to be
non-syntactic. By forcing coreference between the subject and 2-possessor when the
possessor is a 3rd person pronoun, the condition serves an important disambiguating
function in discourse, and it may be that it is not operative when the possibility of
ambiguity in a particular discourse is sufficiently low. If so, the presence of a discourse
context is crucial, because speakers assign bizarre coreferential readings to sentences like
(55) below when they are presented in isolation rather than non-bizarre noncoreferential
readings.
7 John Haviland points out with reference to this text that -be occurs only when the
possessed object is still 'attached' to its possessor, is part of something else, while reference
to the object once separated does not involve -be. He notes that my account fails to do
justice to this correlation. This is correct, but it seems plausible that whether the penis is
attached to the deer or not primarily determines how the narrator will refer to the penis
(whether as "the deer's penis" or as "it" or as "the penis", etc.) and not whether or not to
use PA. Once this choice is made, the possibility or impossibility of PAis determined.
Haviland objects more generally that my account of PAin solely syntactic terms is
inadequate, and that when PAis optional, its optionality is determined by non-syntactic
factors. I am sure this is correct and my attempt to specify necessary and sufficient
syntactic conditions on PAis not a claim that there are no non-syntactic conditions on PA.
H Both Carol Rosen and Paul Postal (personal communication) have pointed out that the
Southern Tiwa data presented in Allen et al. (1984) are consistent with an analysis under
which possessors ascend to 3 from 2-hosts and advance to 2 by 32A, just as in Tzotzil. In
all their examples of I-hosts, the predicate is plausibly unaccusative, making possible an
analysis in which these hosts are 2s. Hence Southern Tiwa appears to be consistent with a
general law limiting PA to structures in which possessors ascend to 3 from 2s. But
languages like Cebuano where Is host PA to I (Bell 1983) are not. A law restricting PA
from 2-hosts to 3 remains a possibility.
APPENDIX TO CHAPTER 8
Possessor Ascension
Animate Possession
Inanimate Possession
(13) Ch- [yj- ik' la jelavel ti s- me7 ti s- malal 7unc:. OCK 168
icp A3 take cl passing by the A3 mother the A3 husband cl
Her husband, would pass by to bring his, mother.
(14) Tz- k'an n070x ch- [yj- elk'an lok'el tal ti y- ajnil 7une. OCK 190
icplA3 want only icp A3 steal away here the A3 wife cls
He, just wanted to steal his, wife away.
1. INTRODUCTION
This chapter deals with the surface constituency of PA clauses and the
interaction with PA of various 'extraction rules', in particular, Focus and
Topicalization. These interactions illuminate properties of the latter two
phenomena.
My goal is to explain two facts. First, that raising of a possessor does
not, apparently paradoxically, preclude its appearing as part of the
nominal constituent from which it is raised. This is shown most clearly by
sentences where the possessor is topicalized as part of its nominal host:
These are cases of PA, since the clauses are all ditransitive, the 3 matches
the 2-possessor in person and number, and the 3 bears no discernible
thematic relation in the clause. The topicalized constituent is the nominal
host from which the possessor is raised, yet the possessor is contained
within it. My proposal is roughly the following: the 2-possessor ascends to
clausal 3, with the consequence that one nominal bears two relations:
2-possessor and clausal 3. The existence of copy ascensions in which an
ascendee is replaced in its earlier relation by a copy pronoun is known
(Joseph 1976, 1978; Perlmutter and Soames 1979). What happens here,
in contrast, is that the ascendee is replaced in its later relation by a
pronoun (an anticopy) which, like other definite pronouns, drops. The
possessor is left to surface within its nominal host, which may topicalize,
as in (1 )-(3).
155
156 CHAPTER 9
There are functional differences between topics and foci. The focus
construction implies the existence of a set of individuals any of which
might have satisfied the sentence, and asserts that only that member of the
set referred to by the focussed constituent in fact does satisfy it. Thus, the
focus constrasts with other (perhaps implicit) elements whose occurrence
in place of the focus would not satisfy the sentence. Foci occur in
prepredicate position. All the following examples are from texts with
enough context included to give a sense of what the focus construction
'means':
(8) "I'll open the cage for you to get out, then. Me, I'll get in" said
Coyote-
Jkobel koyote xa la teo OCK 368
fucking coyote cls there
It was the fucking coyote that was there now.
(9) "I'm planting. I'm planting stones, I'm planting trees," he said,
T(\
(16):
(16)
Ff\
b
(\
Note that topic (Top) and focus (Foe) are taken to be grammatical
relations. What (16) represents is that the focus and the clause in which it
originates make up a constituent, b. The constituent b is the domain for
clitic placement. Whatever this constituent is, it forms a larger constituent
with the topic. Word order rules must place both the topic and the focus
before elements which head neighboring arcs. This fact, together with the
structure in (16), predicts that the topic will precede the focus in sentences
containing both, a correct prediction: in the first example below, vo7on is
topic and ta kok is focus. In the second, 7ovrekon is topic and ta ka7 is
focus.
(17) 7a Ii vo7on -e ta kok ch- 1- 7anilaj. OCK 351
topic the I cl by foot icp B I run
Me, it's on my legs that I run.
160 CHAPTER 9
I now turn to the fact that a possessor can ascend out of its nominal host,
while still surfacing as part of it. Disregarding the topic, the relevant part
of the structure of example (1) (repeated here), is (20):
(1) 7a Ii s- tot Ii Xun -e 7i- j- k'opon -be.
topic the A3 father the Xun cI cp Al speak io
I spoke to Xun's father.
(20)
stot Ii Xune
father
TOPIC, FOCUS, COPY POSSESSOR 161
The possessor ascends to 3, with the result that Ii Xune heads overlapping
arcs: A and B.
In transformational terms, the resulting structure contains two occur-
rences of Ii Xune, one of which must be replaced by a pronoun, leaving
the other as antecedent. The prevailing view in transformational grammar
was probably that the copy would inevitably serve as antecedent, with the
original pronominalized (cf. Joseph (1976, 1978) and Perlmutter and
Soames (1979, pp. 160-3) on copy raising in Greek). Here, however, it is
the copy which is pronominalized, with the original serving as antecedent. 2
In present terms, the issue here is which of the overlapping arcs is
replaced by an anaphoric arc. It is necessary that the 3 be replaced by an
anaphoric arc, leaving Ii Xune to surface as genitive in its nominal host, a
situation represented in (21):
(21 )
3 c2
stot liXune j
father
The anaphoric are C replaces B. Arc C is an ANTICOPY ARC, and its head
an ANTICOPY PRONOUN (see section 6.1 for definitions). The 3 advances
to 2. The final chomeur is stot Ii Xune, a constituent which can be
topicalized, as it is in (1 ):
162 CHAPTER 9
(22)
stot Ii Xune i
father
(25)
Top
3 cz
slibro Ii Xune i
book
(Note that when an interrogative genitive is fronted with its head, the usual
order of genitive and head is reversed, with the genitive (here, k'usi)
phrase-initiaL) An appropriate reply to (28) would be yotal ve7elil
'tortillas for dinner'. The possessor ascends to 3 in which relation it is
replaced by a pronoun. This leaves k'usi to surface as a dependent of the
nominal k'usi yotal which, as the questioned element, precedes the clause
in which it originates. Other similar examples are:
Examples (28) and (30) involve inanimate possession. Example (31) does
not. In (31), the possessor buch'u 'who' raises to 3 and the 3 is replaced
by a pronoun, leaving buch'u to surface within its host. The pronominal 3
advances to 2, putting the host in chomage, and then to 1. The chomeur,
buch'u stot, is questioned.
Two further questions arise. One is whether the overlap in PA clauses can
be resolved by replacing the raised element with a pronoun only when
certain other conditions are met (e.g., when the host is topicalized), or
whether it is always possible. Since I have no evidence on this point, I
assume that it is always possible, with the result that a sentence like (32)
has a surface structure in which sk'ob Ii vinike is a constituent.
(33) a b.
a pronoun;
4.1. Topicalization
Rule (4) accounts for the interpretation of (34):
(34) Ta s- sa7 Ii s- krem -e.
icp A3 seek the A3 son cl
He; looked for his;l'/ son.
TOPIC, FOCUS, COPY POSSESSOR 167
Rule (4) also accounts for the ungrammaticality of (35), since co reference
is impossible between pronouns of different persons:
(35) *Ta j- pak'an y- otaI.
icp Ai make A3 tortilla
(I'm making tortillas for it.)
However, the topicalized versions of (34) and (35) are not subject to (4).
Both (36) and (37) have noncorefential readings. Hence (36) is ambigu-
ous, and (37) is grammatical:
(36) 7a Ii skreme, 7issa7 nan.
Perhaps he; was looking for hisilj son.
(37) 7a Ii yotale, tajpak'an.
l'm making tortillas for it.
The question then is why (36) and (37) are not subject to (4), if (38) is
their structure. Illustrating with (37):
(38)
80
yotal it i tajpak'an
tortillas make
I tajpak'an
make
yotal it
tortillas
The topicalized element is 2 in the basic clause 80 and Top in the overlay
clause 90. The 2 is replaced by an anaphoric (copy) pronoun.
Under this proposal, a clause which hosts topicalization has a stratum
containing a pronoun in place of the topicalized constituent. However, as
now formulated, (4) still applies to (39), since the basic clause in (39) is
monotransitive and contains a 2 with a 3rd person pronominal possessor
(i.e., (4) does not require that the 2 be in any particular stratum.)
However, if (4) is revised so that it holds only if the last 2 in a
monotransitive clause has a 3rd person pronominal possessor, then it will
not apply to (39). Accordingly, (4) should be revised to (40):
(40) Tzotzil 1/2Gen Coreference Rule (final informal version):
If Cn is the last transitive stratum in a monotransitive clause b
and the 2 in Cn has a 3rd person pronominal possessor a and
... , then the initial 1 in b reflex-antecedes a.
Rule (40) does not apply to (39), since Cz is the last transitive stratum in
the basic clause, and the 2 in C z is not possessed. Rule (40) does all the
work (4) did in clauses involving no overlay relations (extractions) and will
TOPIC, FOCUS, COPY POSSESSOR 169
(41 )
The last transitive stratum in (41) is c J, and cJ's 2 contains a 3rd person
pronominal possessor. But co reference between 1 and 2-possessor
always precludes passivization, with the result that (41) is ill-formed
(see discussion in chapter 8).
A 'base' analysis of topicalization under which the 2 is a pronoun in all
strata would also explain why (36) and (37) are not subject to (4). But it
fails because of sentences like (42) and (43) which involve topicalization
and P A. If the initial 2 were a pronoun, PA would be impossible, for PA
is possible only when the 2 is possessed (see chapter 8, section 9).
(42) 7a Ii skreme, 7issa7be nan.
Perhaps hei was looking for hisj!*i son.
(43) 7a Ii yotale, tajpak'anbe.
I'm making tortillas for it.
170 CHAPTER 9
Example (42), which involves both PA and topicalization, has the struc-
ture in (44):
(44)
7issa7be
skrem seek
son
The 2-possessor ascends to 3. How the resulting overlap is resolved is
irrelevant - for concreteness, I assume the Gen arc is erased. The
ascended 3 advances to 2, putting the host in chomage. The chomeur host
is topicalized, determining a pronoun as required by the conditions on
Tzotzil topicalization. Structure (44) is subject to the co reference con-
straint on PA, ruling out the possibility of reflexive coreference between
the 1 and 2-possessor in (42).
In short, PA is possible in (42) and (43) because the clause contains a
possessed 2. However, sentences like (36) and (37) nonetheless escape
(40) - the l/2Gen Coreference Rule - because topicalization makes the
last and final 2 in the clause a pronoun.
4.2. Focus
Rule (40) holds in clauses which host 2 focus. Neither (34) (repeated
below) nor its focussed version (6) (repeated below) can be interpreted
without co reference: 4
TOPIC, FOCUS, COPY POSSESSOR 171
skrem tassa7
son seek
172 CHAPTER 9
(46) b.
yotal tajpak'an
tortillas make
In both cases, the last transitive stratum is c I' and in both, the 2 in that
stratum is possessed by a 3rd person pronoun. Rule (40) then accounts
for the fact that (6) has only the coreferential interpretation, and for the
fact that (45) is ungrammatical. In the latter case, the ungrammaticality is
due to the fact that a 1st person pronoun cannot reflex-antececle a 3rd
person pronoun.
In summary, these contrasts between topicalization and focus follow if
the topicalized constituent is replaced in its clause with a pronoun and the
focussed constituent is not.
The Successor Erase Law requires that any predecessor which is not
erased by a replacer be erased by its successor (see chapter 2, section 2.2,
andJ &P,p.113):
(47) The Successor Erase Law:
If A is B's successor and there is no C which replaces B, then
A erases B.
In Tzotzil PA, the Gen arc has a 3 arc successor. The Successor Erase
Law allows two outcomes for P A constructions. In neither can the genitive
surface qua genitive, for in both the Gen arc is erased. The Gen arc may
be replaced by a pronominal arc (which erases it) or it is erased by its
successor. The Successor Erase Law rules out all outcomes in which the
Gen arc is not erased: one in which the Gen arc erases its successor, and
one in which the successor is replaced by a pronominal arc and is
therefore erased by it. It is the last mentioned outcome which is found in
Tzotzil. Hence the Successor Erase Law is too strong.
Frantz (1979) discusses a similar case. In Blackfoot clauses involving
subject-to-object raising, the complement 1 may surface in the comple-
ment clause with the raised 2 a pronoun. In APG terms, the complement 1
arc has a 2 arc foreign successor, but is not erased by its successor. Since
it is not erased by a replacer either, the result violates the Successor Erase
Law. Instead, the successor is replaced by an anaphoric arc seconded
by the complement 1 arc. Again, this configuration is in line with
pronominalization possibilities in Blackfoot: a complement 1 can antecede
a main clause 2 when the two are in a coreference relation, i.e., when the
relations involved are both initial stratum relations. Hence, Blackfoot
pronominalization constraints, like those of Tzotzil, do not distinguish
coreferential pronouns from copy pronouns. Or, in terms of arcs, they do
not distinguish overlap involving initial arcs from that involving predeces-
sor/successor pairs.
If the Successor Erase Law is abandoned entirely, then except for cases
in which the predecessor is replaced, there will be no constraints on the
174 CHAPTER 9
John like
The 2 arc A has a Top arc foreign successor, which erases it. Arc B is also
sponsored by its companion arc, C. It is not clear what sponsors the
companion arc; I leave this open.
6.5. Conclusion
The two copy structures discussed here, possessor ascension and topi-
calization, require little language-specific stipulation. A language-specific
rule is needed when a successor structure must be limited to the plain or
copy (or anticopy) type. In the case of PA, there is no need to limit it to
the anti copy structure, since allowing copy and plain versions has no
known undesirable consequences. Hence, no rule is needed here. Topi-
calization, on the other hand, must be limited to a copy structure and
focus to a plain structure. This is accomplished by rule (53). Not
surprisingly, rule (53) has the same general form as rules (42)-(43) of
chapter 6 which determine plain and reflexive (copy) unaccusative
structures for particular predicates. The facts discussed in this chapter all
follow from rule (53) and the revision of the co reference condition in
monotransitive clauses (51), given the theory of erasure assumed here.
Finally, allowing anticopy structures for possessor ascension resolves
the paradox posed by the apparent discrepancy between the relational and
constituent strw::ture of PA sentences. A discrepancy of a similar sort is
discussed in the following chapter. There, elements which are clearly not
surface constituents nonetheless control agreement and antecede r~flexives
as a unit. The solution proposed there is quite different from the one
proposed for PA, and depends on the multistratal character of APG.
NOTES
I However, PPs of certain types occur in the same structural position as topics. Because of
the position of the clitic in the following example (see below in the text), the fronted PP
cannot be a focus:
(i) Ta primero 7ak'ubal ch'abal la, ta xchibal 7ak'ubal ch'abal
on first night nothing cl on second night nothing
la. OCK 370
c/
On the first night, there was nothing, on the second night there was nothing
It seems likely that only PPs which function semantically as sentential adverbs can be
fronted in this way.
2 This conception is incompatible with the also prevailing view that there is no rule of
pronominalization.
3 Aissen (1979) argued that the possibility of placing PPs between elements corresponding
to sk'ob and Ii vinike was evidence for their nonconstituency, as, for example, in:
(ii) 7i- k'as -b -at x- chak ta te7 ti bolome. OCK 45
cp break io psv A3 ass with stick the jaguar
The jaguar's ass was. broken with a stick.
However, the grammaticality of (ii) entails the desired conclusion only if PPs cannot
separate a head noun from its possessor in cases where PAis not involved. While such
178 CHAPTER 9
examples are not easily constructed, I am not prepared to say now that they do not exist. If
they do, then (ii) is irrelevant to the question of constituency.
More suggestive is the fact that the nominal which is final chomeur can be questioned
in clauses involving PA:
(iii) K'usi a- k'as -be Ii Petul -e?
what A2 break io the Petu! cl
What of Petul's did you break?
presented out of context. John Haviland has suggested that in particular contexts, (6) may
be interpretable without coreference. In view of chapter 8, footnote 6, this would be
entirely expected.
5 32A would be impossible because the resulting structure would violate the Unique
Eraser Law (chapter 2, section 2), which limits the number of erasers of an arc to one. In
32A, the 3 arc is erased by its 2 arc successor, bringing the number to two if the 3 arc is
also erased by its Gen arc predecessor.
Nothing in Frantz's discussion makes clear whether the predecessor can erase its
foreign successor. The raised element controls agreement on the main verb, showing that it
TOPIC, FOCUS, COPY POSSESSOR 179
must be final 2 in the main clause. But since all foreign erased arcs are final arcs, erasure
by its (foreign) predecessor would not preclude agreement.
n As noted by Paul Postal (personal communication), (50) may prove to be a theorem
from other principles given the existence of the Internal Survivor Law (1 & P, p. 526),
which insures that one of a predecessor/successor will be erased, thereby guaranteeing part
of what (50) guarantees.
7 The Revised Successor Erase Law would allow an overlay arc to be replaced by an
anticopy arc seconded by the predecessor arc. In general, this seems not to happen, and it
certainly must be blocked in Tzotzil topic constructions where the topic cannot be a
pronoun anteceded by some nominal in the following basic clause. At worst, a Tzotzil rule
which prohibits replacement of an overlay arc by an anticopy arc is required.
SURROGATE AGREEMENT
1. INTRODUCTION
2. POSSESSOR ASCENSION
In all these examples, -ik must cross-reference the final absolutive, and
cannot cross-reference the final chomeur. In contrast, consider a sentence
like (6), which does involve PA.
(6) L- i- s- k'el -be -ik j- ch'amaltak Ii Xun -e.
cp BI A3 watch io pI Al children the Xun cl
Xun watched my children.
182 CHAPTER 10
(7)
Cho
lisk'elbeik
watch
H
Gen
Ii Xune
jch'amaltak I
children
chomeur hosts the ascension of a PTAC (i.e., the final absolutive), while in
(5), the chomeur does not.
My claim is that number agreement in Tzotzil can be controlled either
by a PT AC or by a constituent which hosts the ascension of aPT AC.
In contrast to number agreement, person agreement can only be
controlled by a final 1 or final 2. In particular, it cannot be controlled by a
PA host. If person agreement in (6) were controlled by the host, the verb
would agree with a 3rd person absolutive (and therefore bear no overt B
affix). The result is grammatical, but does not involve PA:
(9) 7i- s- k'el -be -ik j- ch'amaltak.
cp A3 watch io pi Al children
He/she/they watched my children for them. or
They watched my children for him/her/them.
Here the 3 is thematic benefactive. No PA is involved, and -ik must cross-
reference a final term, either 1 or 2.
The anomalous agreement in PA structures is interesting evidence for
the ascension analysis. If the chomeur and the final 2 were not relevantly
connected, there would be no explanation for the fact that the former can
be a surrogate agreement controller. Of course, they are also connected in
that the final 2 puts the chomeur in chomage, but this is not the relevant
connection, since surrogate agreement is possible only in PA structures,
and not in other ditransitive structures (much less in passive structures,
which also involve chomage). Further, I argue in section 3 that it is a
general principle of Tzotzil that ascension hosts can control number
agreement, further supporting the ascension analysis of PA structures.
3. CONJUNCT UNION
In (10), the verb agrees with an element which is neither a final 1 nor a
final 2:
(10) 7i- bat -ik xchi7uk y- ajnil Ii Xun -e.
cp go pi with A3 wife the Xun c/
Xun went with his wife.
The final 1 is Ii Xune, but the verb is plural, agreeing, in some sense, with
the plurality of the set of 'goers', a set which includes Xun and his wife. An
understanding of sentences like (10) requires some discussion of xchi7uk,
the word translated 'with' above.
3.1. -chi7uk
Xchi7uk, as well as the other elements discussed in this section, is based
184 CHAPTER 10
p Adv
P 2
live7 I
ate
jchi7uk him
with
The interest of these examples lies in the fact that they appear to have
singular subjects (vo7on 'I' in both cases), but plural verbs (1st person
plural exclusive in both cases). The nominal flagged by xchi7uk names the
person who, together with the subject, makes up the plurality marked in
the verb. Examples (29) and (30) do not involve any of the constructions
discussed earlier. Xchi7uk is not predicate of a subordinate clause
because it does not agree with the nominal which would be its subject, the
1st person pronoun. Nor can it be a conjunction because it does not
conjoin two overt conjuncts. A similar example is:
Here, the direct object is the 2nd person pronoun vo7ot, with xchi7uk
again flagging the nominal which, together with the direct object, deter-
mines the plurality marked in the verb. Xchi7uk is not a predicate since it
does not agree with vo7ot, which would presumably be its subject. Nor is
it a conjunction since that would require something like avajnil xchi7uk
vo7ot with xchi7uk between the conjuncts. In need of explanation then is
that the verbs in all three examples are plural, while the expected
controllers (final Is in (29)-(30) and final 2 in (31)) are singular. I
propose that sentences like (29)-(31) involve ascension out of coordinate
structures with agreement controlled by the ascension host. Under this
analysis, the initial 1 of (29) would be roughly the coordination corre-
sponding to 'me and our compadre R.T.' Vo7on 'I' ascends and takes on
the grammatical relation of the host nominal. In this respect, conjunct
union is like other ascension constructions. The other conjunct cannot
remain a conjunct, so it too ascends into the clause. It is not obvious what
relation is borne by this ascendee, but it clearly cannot take on the relation
of its host, as this would violate stratal uniqueness. I propose that it
ascends as a DEAD NOMINAL. The rationale for this is discussed in more
detail in section 3, but the basic idea is this. Conjunct union shares with
other union constructions the property that all dependents of the lower
node ascend into the main clause. This creates the opportunity for
potential violations of stratal uniqueness, violations which are avoided
through the Dead relation (see chapter 11). Conjunct union appears to be
analogous. I will refer to nominals which ascend out of coordinate
structures as dead nominals as DEAD CONJUNCTS. Dead conjuncts are
flagged with xchi7uk. What I have sketched so far is represented below
(the grammatical relation borne by a conjunct to the coordinate structure
is Conjunct (Con):-
SURROGATE AGREEMENT 189
(32)
Ii 7ayotikotik
went Con
Con
follow the 1. This prediction is correct. Xchi7uk yoltak and Ii jtamole can
be switched in (34), retaining grammaticality. The same is true in (31) and
(33).
Conjunct union occurs when the conjuncts are not of equal discourse
prominence, prominence being in part determined by the hierarchy:
(35) 1st person
2nd person
3rd person
The more prominent conjunct ascends and takes the grammatical relation
of its host, leaving the less prominent one to ascend as dead. Thus, in (29),
the 1st person conjunct ascends as 1, the 3rd person conjunct as dead. In
(31), the 2nd person conjunct ascends as 2, the 3rd person conjunct as
dead. In (37) below, a 1st person conjunct ascends as 1, leaving the 2nd
person conjunct to ascend as dead. In this respect, conjunct union is like
other advancement and ascension rules which, as is well-known, often
promote discourse prominent nominals to central syntactic relations (see
Partee (1971 )).
The conjunct union analysis claims that there are sentences containing an
underlying coordinate structure whose conjuncts, while forming a con-
stituent in the initial stratum, do not form one in surface structure.
That these elements do not form a surface constituent is clear. In (31),
for example, there is no plausible analysis which would make the string
xchi7uk avajnil Ii vo7ote a constituent. It is not a coordinate structure,
nor is it a head-modifier structure with xchi7uk avajnil modifying Ii
vo7ote ('you with your wife'). That too would require the other order.
Further, if xchi7uk avajnil Ii vo7ote were a constituent, it should be
possible to topicalize it. Example (36) shows that it cannot be topicalized.
(36) *7a Ii xchi7uk avajnilli v070te ...
Turn now to the claim that the nominals in question form a coordinate
structure at some level. This is supported by the existence of sentences
which involve conjunct union and contain a reflexive nominal, as in (37).
Sentence (37) means literally something like 'I will marry each other with
you', and contains a reflexive nominal understood reciprocally:
(37) 7a ti mi k- ik' j- ba -tik xchi7uk
topic the ? Al marry Al self A *Iplinc with
v070te ... SSS 147
you
If you and 1 get married ...
SURROGATE AGREEMENT 191
90 kik'
marry
Con Con
vo7ot vo7on
you I
(39)
90
[1 pi inc.] kik'
pronoun marry
Con
vo7ot vo7on
you I
Gen
H
kik'
90 [1 pi inc./ marry
pronoun
jbatik
Con Con each other
vo7ot vo7on
you I
SURROGATE AGREEMENT 193
The reflexive nominal jbatik in (37) agrees with a 1st person plural
inclusive genitive, exactly as predicted by (40). This accounts for the
reflexive syntax as well as the inflection on the reflexive nominal. The
reflexive rule developed in chapter 5 accounts for these cases without any
modification.
Turning now to conjunct union, the initial stratum 1 in (40) is a
coordinate nominal. 'I' ascends with the grammatical relation of its host 1,
leaving the other conjunct to ascend as dead, and to be flagged by
xchi7uk. The full structure of (37) then is (41).
(41 )
Dead C2• 3 1 C1 2 c1
@ ® 1 C 2,3
kik'
I marry
90 [1 pi inc.]
pronoun
jbatik
each other
v070t v070n
you I
As (41) makes clear, the initial and final strata of (37) are significantly
different. While both strata are transitive, the initial and final Is are
different, as are the initial and final 2s. The coordinate nominal is both
initial 1 and initial 2. But, the final 1 is the conjunct vo7on '1', and the final
2 is the reflexive nominal jbatik. Example (37) simultaneously provides
evidence for an underlying coordination and for the surface non-
constituency of the elements which make up that earlier coordination.
As such, (37) is very strong evidence for the conjunct union analysis.7
Two questions arise regarding control of agreement in conjunct union
structures. One is whether the predicate can agree in number with the final
1 at all when the host is available as a controller. And if it cannot, a
second question is whether agreement with the host is obligatory. The
194 CHAPTER 10
In one other case the predicate agrees in number with an element which is
neither final 1 nor final 2. Consider first (44), an intransitive clause in
which the subject is singular, but the verb is inflected as plural:
(44) 7i- bat -ik Ii Xun -e.
cp go pi the Xun cl
Xun went (with someone).
The plural suffix in (44) (and analogously in all like examples) entails that
Xun went with someone else, i.e., that the set of goers included at least
two people: Sentences (45)-(47) exemplify the same construction. The
first contains a non-verbal predicate:
(45) Tey -ik ta ch'ivit Ii Xun -e.
there pi at market the Xun cl
Xun is at the market (with someone).
In (46), the plurality cross-referenced on the verb is associated with the 2,
and in (47), with the (initial) 3.
(46) 7i- k- il -ik ta ch'ivit Ii tottik Xun -e.
cp A 1 see pi at market the Mr. Xun cl
I saw Mr. Xun at the market (with someone).
nominals, since the individuals they name are members of the set cross-
referenced by the plural suffix.
I propose that these sentences involve conjunct union, with the
indefinite nominal UN being one of the conjuncts. The initial subject of
(44) then corresponds to 'X un and UN'. Xun ascends as subject, leaving
UN to ascend as dead. The structure of (44) is:
(48)
Con
Con
UN Xun 7ibatik
went
4. SUMMARY
revised: the predicate optionally agrees in number with its final 1, or its
final 2, or with a constituent which hosts the ascension of its final 1 or
final 2.
This section has two goals. The first is to establish general properties of
conjunct union, expressed in a proposed Conjunct Union Law (rule (53)
below). The second is to sketch a theory of agreement control which is
consistent both with the agreement laws of chapter 3 and with the facts of
Tzotzil surrogate agreement just discussed. Agreement in conjunct union
structures appears to be inconsistent with the Controller Agreement Law
(p. 56), which requires that controllers head final arcs, for in conjunct
union, the controller heads no final arc. Agreement in PA structures can
be made consistent with the Controller Agreement Law by allowing a
chomeur to control agreement, but this sheds no light whatever on Tzotzil
agreement since chomeurs do not in general control agreement. To sustain
an insightful account of these facts which is consistent with the Controller
Agreement Law, I propose that under certain conditions, nominals may
pass their features to other nominals. When the latter are lawful agreement
controllers, features of unlawful controllers may be realized on the
predicate. The agreement controllers in Tzotzil PA and conjunct union
structures satisfy the conditions for feature passing.
Con
Con
198 CHAPTER 10
In the set of foreign successors T, one member has the R-sign of its local
sponsor, a fact which would follow from the Relational Succession Law, as
currently stated. The Relational Succession Law requires that a raised
nominal assume the grammatical relation of its host. However, the
Relational Succession Law predicts that all members of T have the R-sign
of their local sponsor, a prediction which is both factually false and one
which contradicts the Stratal Uniqueness Law. It is obviously desirable
that the Relational Succession Law apply to that unique foreign successor
which does have the R-sign of its local sponsor (i.e., its support), since the
existence of such an arc is presumably not accidental. Hence, either that
law must be modified to allow for foreign successors which do not have
the R-sign of the local sponsor, or an analysis of conjunct union must be
provided under which these latter foreign successors are not locally
sponsored by their support. Both approaches are possible, and it is not
clear at present how to motivate a choice. I take the first approach and
propose a revision of the Relational Succession Law which allows foreign
successors which have R-signs different from that of the local sponsor. 8
The formulation of the APG Relational Succession law (1 & P, p. 710) is
equivalent to (50).
(50) APG Relational Succession Law:
If A is a Term x arc and local sponsor of a nominal arc Band
B is C's foreign successor and C is not a Gen arc then B is a
Term, arc.
The Relational Succession Law can be made consistent both with the
cases it was originally intended to explain and with conjunct union by
requiring not that every foreign successor B be a Term x arc, but that some
foreign successor be a Term x arc. Since ascensions like subject raising
involve only a single foreign successor, these are covered as they were
under the original formulation. In cases like conjunct union which involve
more than one foreign successor, at least one Term x foreign successor is
guaranteed, but non-Term x successors are allowed as well. The Stratal
Uniqueness Law will then limit the number of Term x foreign successors to
one, and principles perhaps peculiar to conjunct union will determine the
R-signs of other foreign successors. In (51) a revision of the Relational
Succession Law is proposed.
(51) Revised Relational Succession Law:
Let T = set of nominal arcs. Then: For all B in T, if A is a
Term x local sponsor of Band B is C's foreign successor where
C is not a Gen are, then there is some B in T which is a Term x
arc.
Turn now to the R-sign of non-Term x foreign successors in conjunct
union structures. Possible R-signs include: an oblique R-sign (perhaps
SURROGATE AGREEMENT 199
Comitative), Cho(meur), Con, Dead, or some new relation. The first two
are ruled out by current APG laws. To posit an oblique relation would
violate the Oblique Law (PN 49) which requires that no oblique arc be a
successor. The chomeur relation makes some sense here because the
chomeur relation exists to avoid violations of the Stratal Uniqueness Law,
but accommodating this case would require a revision of the Chomeur
Law, since the predecessor arc is not overrun. To posit Con here is to
allow Con arcs outside of coordinate structures, which is presumably
undesirable. Of previously recognized R-signs, this leaves Dead.
J & P restrict the Dead relation to union constructions, assigning it
to those nominals other than the 1 and 2 which are raised from the
complement into the main clause (but see chapter 11). In the R-sign
typology of J & P (page 23 of chapter 2 above), Cho and Dead arcs
constitute the class of Derivative arcs, a subset of Central arcs. Like
other Central R-signs, Cho and Dead are nominal R-signs and occur in
basic clauses. Unlike other Central R-signs, they are never associated with
initial stratum arcs. Cho arc successors are always local, while Dead arc
successors are always foreign. 9 Accordingly, I propose that all foreign
successors in conjunct union structures are Dead arcs except for the one
whose R-sign is determined by that of its local sponsor:
(52)
Con
R-predecessor (see chapter 2, page 31), both Con arcs and arcs with Con
arc predecessors have Con arc R-predecessors. This allows rule (55):
(55) Tzotzil xchi7uk Flagging Rule:
If A has a lower pioneer successor whose companion arc is
headed by xchi7uk, then A has a Con arc R-predecessor or
The disjunction "or ... " allows xchi7uk other functions besides conjunct
flag. I 2
Putting all this together, the structure of (29) is (56). Sentence (29) is
repeated below with irrelevant parts omitted.
(29) L- i- 7ay -otikotik xchi7uk jkumparetik Romin Terato\.
cp BI go Blplexc with our-compadre R.T
I went with our compadre Romin Terato\.
(56)
li7ayotikotik 1 jkumparetik
went Romin Teratol xchi7uk
our compadre R. T with
Stratal Uniqueness Law. Tzotzil rule (54) requires that vo7on ascend as 1
because it outranks jkumparetik R.T. on the person hierarchy. In terms of
arcs, A's foreign successor is a 1 arc, like A's local sponsor, E. And B's
foreign successor is a Dead arc. The flagging of the dead conjunct is
represented by F and its branches. The key fact is that jkumparetik R.T.
can be flagged with xchi7uk because it heads a Con arc. I return to the
agreement in (29) below.
with
The coordinate structure 90 heads both the initial 1 and 2 arcs, predicting
the coreferential interpretation of (37). Node 90 itself is 1st person plural
inclusive since it governs the 1st and 2nd person pronouns. The initial 1
and 2 arcs A and B overlap, sharing the coordinate node 90 as head.
Node 90 governs the 1st and 2nd person pronouns. Arcs A and B
cosponsor C, which is an anaphoric replacer for B, headed by the 1st
person plural inclusive pronoun, this being the pronoun that agrees in
SURROGATE AGREEMENT 203
person and number with B's head, 90. Arc C is a reflexive arc,
anaphorically connected to its seconder, A. As a 2 arc anaphorically
connected to a neighboring 1 arc, C satisfies the conditions of the reflexive
rule (chapter 5, rule (19)), and is replaced by a camouflage arc, D, whose
branches are C's lower pioneer successor and its companion H arc. At the
same time, each of 90's co limbs, F and G, has a foreign successor, defining
(57) as a conjunct union structure. Since F's head outranks G's, F has a
Term arc successor, a 1 arc, as determined by the Relational Succession
Law. It overruns its local sponsor A, and erases it. Arc G has a Dead arc
foreign successor, I, which in turn has a lower pioneer successor M. Arc I
is replaced and erased by J, closure arc for the lower pioneer structure.
The predicate agrees in person with its finalIst person subject. It is not
inflected for number. Jbatik agrees with its genitive in person and number.
The clause represented in (57) governs only three surface arcs: the P arc
E, the 2 arc D and the Dead arc J. These represent the three constituents
of (37). Tzotzil word order requires that the predicate be clause-initial and
that the reflexive nominal immediately follow the predicate.
seem
there unicorns
Together, the Lateral Feature Passing Law and the Tzotzil Lateral Feature
Passing Rule restrict the class of surrogate controllers in Tzotzil to
ascension hosts in PA and conjunct union constructions. Under this
analysis, all the agreement laws proposed in chapter 3 can remain, as can
all Tzotzil agreement rules proposed there.
As illustration, consider (6), repeated below, in which jch'amaltak, final
chomeur, appears to control number agreement.
(62)
80
"-
H "-
St '" \
\
\
\
jch'amaltak 1 -ik -i- -s- k'el
children [3 plj [Blj [A3j watch
Structure (62) involves PA as the Gen arc E has a 3 arc foreign successor
B. Arc B has a 2 arc local successor D which overruns A (i.e., (62)
involves 32A). As final Erg and Abs arcs, F and D sponsor set A and set
B agreement arcs, respectively. The plural node 80 is ultimately respon-
sible for the plural inflection on the verb. Under the present proposal
though, neither of the arcs which have 80 as head (namely, A and C)
sponsors the plural agreement arc, because neither is a possible agreement
sponsor in Tzotzil. Arc A is not a final arc, and C is not aI, 2, or Gen
arc. Instead, the plural arc is sponsored by D, whose head is plural, though
non-inherently so, through agreement with A's head. Arc A's head can
pass its features to D's head because A and D satisfy relevant constraints.
Arc D overruns A (Lateral Feature Passing Law). And A locally sponsors
an R-predecessor of D, namely B. That is, A's head hosts the ascension of
D's head (Tzotzil Lateral Feature Passing Rule).
The two sets of features must be kept distinct; they cannot be unified in
the sense of Gazdar et al. (1985). The result of unifying the categories
associated with 80 in (62) with those associated with D's head would be
/1 plj. But the Af arc that D sponsors is headed not by a 11 plj affix, but a
[3plj affix. The [Blplj affix would be -otikotik (exc) or -otik (inc). Hence
under the pre.sent account, D's head must be associated with two distinct
sets of categories, /1sgj and [3plj, with the 3pl-Af arc sponsored by virtue
of the latter, and the B I-Af by virtue of the former.
SURROGATE AGREEMENT 207
(63)
,/
/
I
/
I
/
I jkumparetik RT
t our compadre R. T.
7ay -1- -otikotik xchi7uk
go [B1] [B1 pi exc.] with
6. CONCLUSION
Taken together with the proposed laws, and rules established in earlier
chapters, the four rules proposed in section 5 account for Tzotzil conjunct
union and surrogate agreement. Two rules, (54) and (55), account for
language-specific properties of conjunct union: the fact that ascension is
determined by a person hierarchy and the form of flagging for dead
conjuncts. All other aspects of conjunct union are determined by (53), the
Conjunct Union Law. Two rules account for language-specific properties
of surrogate agreement. One is (61), which restricts surrogate agreement
in Tzotzil to cases involving ascension. The other, only alluded to, restricts
Tzotzil surrogate agreement to cases of number agreement.
The account of Tzotzil surrogate agreement is important to Tzotzil
grammar. But from the viewpoint of the theory of agreement control
proposed in section 5.5, this account is simply a working out of the
consequences of that theory for a particular language. The Controller
Agreement Law of chapter 3 and the Lateral Feature Passing Law of
this chapter ((60», are intended to have applicability beyond Tzotzil.
Surrogate agreement is attested in many languages (see footnote 13), and
these laws make predictions about the structures in which it is found.
NOTES
I Example (I) has a grammatical reading on which the recipient is understood to be a set
of individuals one of whom is Maruch. Example (2) has an analogous grammatical reading.
These readings are explicated below (see section 3.4).
2 Kol is not grammaticality plural (its plural is koltak). This is compatible with the claim
that the plural verb agrees in number with kol if agreement requires compatibility of the
relevant categories of the elements involved in agreement, rather than identity (see chapter
3, note 7). I assume that kol is neither singular nor plural, and hence can be cross-
referenced either by a singular or plural verb.
In principle, -ik could cross-reference the subject in (X) but this is unlikely here. For
one thing, it is this suffix which forces the translation 'children' for kol, a nominal which is
not grammatically plural. Further, at this point in the story the narrator has in mind a
group of children, mentioned two lines earlier as koltak 'children', and the reference in (8)
is to the same group. Nor does the narrator appear to have in mind a plural entity for
subject. The sentencc citcd in (8) is preceded by two syntactically similar clauses Cthere is
no one who ... '), and in neither does the verb cross-reference a plural subject.
.1 Historically, xchi7uk probably bears the A3 prefix, S-, assimilated to x- (see Phono-
logical Rules 4).
4 The reader will notice that -chi7uk is not clause-initial in any of these examples. What
The subject of achi7uk is vo7ot 'you (singular)', and its direct object is Ii jkumalee 'my
comadre'. The verb labatik, however, agrees with a 2nd person plural subject (corre-
sponding to 'you and your wife'). Therefore, the subject of -chi7uk is not, strictly speaking,
coreferential with any nominal in the main clause, though its reference is included in the
reference of the main clause subject.
fi The protasis of a conditional is generally presented as a topic.
7 Strikingly similar facts are described for Navajo in Hale (1975), a paper brought to my
successor which has the R-sign of its support can be locally sponsored by that support,
with the others having some distinct local sponsor. One possibility (suggested by Paul
Postal (personal communication)) is that the privileged successor erases its local sponsor
(its predecessor's support) and then locally sponsors the foreign successors of all remaining
Con arcs.
9 Gibson and Raposo (1986) propose to enlarge the class of structures in which Cho arcs
may appear, allowing, in APG terms, Cho arc foreign successors. At the same time, they
restrict the class of R-signs by eliminating Dead arcs.
10 Three laws in J & P mention Dead arcs, two of which require some revision if Dead
arcs are extended to conjunct union. PN Law 72 requires (roughly) that all Dead arcs are
foreign successors of Central arcs. Since Con arcs are not Central arcs (see chapter 2, (4)),
this law must be changed to allow Con arc predecessors for Dead arcs. PN Law 73
restricts Dead arcs to clause union structures, and must be modified to allow for their
occurrence in conjunct union structures. Conjunct union and clause union have this in
common: if A is the predecessor of a Dead arc B, then all of A's final stratum neighbors
have foreign erasers which are neighbors of B. In causative clause union, all dependents of
the complement are raised into the main clause, and in conjunct union, all dependents of
the coordinate structure are raised. In APG terms, all final arcs in the subordinate
structure are foreign erased by successors. In so-called Equi union (see Aissen and
Perlmutter (1983) and J & P (chapter 8.6)), the Equi victim is not raised, In APG terms,
the final complement arc it heads is foreign erased by a non-successor, and all other final
arcs are erased by foreign successors. This might serve as the basis for a revision of PN
Law 73 along the following lines:
(i) If A is an organic Dead arc with a predecessor B and a local sponsor C, then
every final neighbor of B is foreign erased by a neighbor of C.
(Organic arcs have overlapping sponsors. The restriction to organic Dead arcs in (i) allows
Dead Marquee Closures, like F in (56).)
210 CHAPTER 10
PN Law 71 governs the appearance of Dead arcs in clause union structures, and is
unaffected by the extension oi Dead arcs to conjunct union.
II I say "In part" because the person hierarchy is irrelevant when both conjuncts are 3rd
person.
12 Rule (55) does not require flagging. This is correct for conjuncts, where the flag is
optional:
However, dead nominals in conjunct union do require flags, so (55) by itself allows for
possibilities which do not occur. I assume the existence of a more general constraint which
allows only certain clausal dependents (e.g., final Is, 2s, 2 Chos) to be unflagged. Since
dead conjuncts are clausal dependents, they must be flagged. Conjuncts, on the other hand,
are not clausal dependents, and thus are not subject to this restriction.
13 Aissen (to appear) argues that this condition is relevant to agreement systems in
communication).
15 This rule is deficient in two respects. It is necessary to exclude the heads of anaphoric
arcs from the domain of the rule. The head of an anaphoric arc A agrees with the head of
A's seconder even though no overrun is involved. It is also necessary to exclude cases of
vertical feature passing involving the head of a phrase and the phrase itself, i.e., the
phenomenon covered by the Head Feature Convention of Gazdar et al. (1985). Such cases
will not, in present terms. involve overrun either.
10 The caveats of note 15 apply here.
APPENDIX TO CHAPTER 10
(4) 7i- k'ot -ik xchi7uk s- tzeb j- bol Mat yo. SSS 159
cp arrive pi with A3 daughter A I bro-in-Iaw Maryo
She arrived with my brother-in-law, Mat yo's, daughter.
1. INTRODUCTION
When the verb 7ak' (which in other contexts means 'give' or 'put') takes a
clausal 2, it has a permissive or causative sense. The following examples
do not involve clause union: the complement clause is both initial and
final 2 of 7ak'. This is clear from the fact that 7ak' has a 2 which is
invariably 3rd person (it bears no overt set B affix). The complement
clauses in the first two examples are intransitive; those in the second two
are transitive. The predicate of the complement clause is in the subjunc-
tive. Hence, it is not inflected for aspect, and when intransitive, it is
suffixed with -uk (-ik- word-internally). Because the verb bears no aspect
prefix, set B suffixes are used.
(4) Mu x- [yj- ak' ve7 -ik -on. OCK 380
not nt A3 let eat subj Blsg
It doesn't Jet me eat.
you 7ak'
let
kuch' I vo7
drink water
CLAUSE UNIONS 215
In (10), the main verb is suffixed with -be, while in (9), it is not.
Moreover, the main verb in (10) must be suffixed with -be, while that in
(9) cannot be.
b. *L-i-y-ak'-be kom-ik-on.
(15)
he
komikon
stay
Xun liyak'be
let
jtuch'
cut
CLAUSE UNIONS 217
The clearest argument that union causatives involve two initially distinct
clauses is the fact that both verbs cross-reference dependent nominals. In
particular, the complement predicate (the element which bears the union
relation in the union clause) cross-references its final 1 and 2. So, kom-
ik-on in (14a) is suffixed with Blsg, agreeing with the final (absolutive) 1
of the complement clause. Jtuch' in (13a) bears an A 1 prefix, agreeing
with the final (ergative) 1 of the complement clause. Since Tzotzil
predicates agree with their final 1s and final 2s, this agreement pattern is
the expected one in union constructions. Failure of agreement would
require additional stipulation.
There arc three arguments that the complement 1 is raised into the main
clause in causative clause union. The complement 1 is 2 in the union
clause whether the complement clause is transitive or intransitive. (In the
case of transitive complements, of course, it raises to 3, and advances to
2.) The most obvious argument, just mentioned, involves agreement.
These facts are regular if such sentences involve PA in the union clause.
The complement clause in all three examples is intransitive, so that its 1 is
raised to 2. Once raised, it is in a position to host the ascension of its
possessor to 3. Note that the ascension host is a non-initial 2. This 3
advances to 2, accounting for the presence of -be on the main verb. As
final absolutive, the ascended and advanced possessor controls set B
agreement on the main verb. In (23), for example, the verb agrees with a
final 2nd person absolutive. Sentence (23) has the RN in (26):
(26)
chakak'be
let
H Gen
latzebe you
daughter
I let your daughter go.
Examples (23)-(25) contrast sharply with the ill-formed (11) and (14b),
though these are also cases in which the complement is intransitive and
the main verb is suffixed with -be. The crucial point is that in the latter
cases, the complement 1 is not possessed and therefore does not host PA
once it is raised into the union clause. The union clause can thus contain
no 3, thus can involve no 32A, and therefore cannot contain -be.
In accord with the coreference condition established in chapter 8,
section 2, PA is possible only whel1 the 1 and 2-possessor are not
reflexive coreferents. If PA is involved in (23)-(25), then coreference
•
220 CHAPTER 11
The evidence cited above (agreement, passive, PA) all argues that the
complement 1 is raised in union constructions. For it is the complement 1
which controls agreement, passivizes, and hosts PA in the union clause.
None of it shows, however, that the complement 2 is raised, and
everything said so far is compatible with an analysis in which only the 1 is
raised.
For concreteness, consider as an alternative to a union analysis the
hypothesis that only the complement 1 is raised. The 1 of an intransitive
complement is raised to 2, while the 1 of a transitive complement is raised
to 3. In the latter case, 32A puts the clausal 2 in chomage. The structures
in (29) represent the subject raising analysis::'
(29)
a. b.
CLAUSE UNIONS 221
Recall that 1st and 2nd person pronouns cannot be put in chomage
(chapter 7, pages 116-7 and rule (55)). The union and subject raising
analyses make different predictions about sentences which translate 'Let
Maruch visit me' or 'Let Maruch wait for us'. The union analysis «30a)
below), predicts that these sentences are ungrammatical because the
complement 2 is final chomeur in the union clause. The subject raising
analysis «30b) below), predicts that they are grammatical, since no rules
are violated. In particular, the nominal which is complement 2 is never put
in chomage:
(30) a. Union
you
let
svula7anon
visit
Maruch
222 CHAPTER 11
(30) b. Raising
svula7anon I Maruch
visit
let
sk'opon
address
224 CHAPTER 11
(35) b. Raising
7ak'bo
let
sk'opon sbaik
address selves
Hence, the union analysis predicts that the relevant sentences are
ungrammatical, while the raising analysis predicts the opposite. Since they
are ungrammatical, the union analysis is again supported:
(36) *7ak' -b -0 s- k'opon s- ba -ik.
let io imp A3 speak A3 self pI
(Let them address each other.)
(37) *7ak'-b-o s-kolta s-ba-ik.
(Let them help each other.)
Compare (36)-(37) with sentences like (38)-(39) which are super-
ficially similar, but involve no reflexive nominal, and therefore cannot
violate constraints involving reflexives.
(38) 7ak' -b -0 s- k'opon s- tot.
let io imp A3 address A3 father
Let him address his father.
Note too that the unreduced counterparts of (36) and (37) are well-
formed. Although they contain reflexives, they involve no violation of
chapter 7, (31 )/(59) since the reflexive is not raised into the main clause
and is not put in chomage.
(40) 7ak' -0 s- k'opon s- ba -ik.
let imp A3 address A3 self pi
Let them address each other.
7ak'
let
CLAUSE UNIONS 227
7ak'
let
Sentence (51) has the partial structure in (53). The same nominal
functions as initial main clause 1 and initial complement clause 2. The
complement 2 is replaced by an anaphoric pronoun which raises to 2 in
the union clause. The union clause 2 is reflex-anteceded by its neighboring
1 and is replaced by the reflexive nominal.
(53)
1
1
7iyak'
let
sba
self
Two facts show that the reflexive nominal is a final dependent in the union
clause. First, it must be a final c1ausemate of the main 1, because reflexive
nominals and their antecedents must be final c1ausemates. Second, it must
be final 2 in the clause whose predicate is 7iyak' because it immediately
follows 7iyak'. A reflexive immediately follows the final predicate in its
clause (see chapter 5, page 78 and chapter 7, page 114). The reflexive
cannot follow the predicate of the complement clause; this verb bears the
union relation in the union clause. Compare (51) with (54):
(54) *7a Ii Maruch-e mu x-[yj-ak' 7il-uk sba.
2.5. Conclusion
now traditional, RG ideas about clause union. In essence, they are that
union structures are initially biclausal but superficially monoclausal. The
fact that predicates agree with their final dependents in Tzotzil provides
the primary evidence that union structures have a distinct complement
clause governing a full set of syntactic dependents, for the complement
verb agrees with its final dependents. A variety of facts show that
dependents of the complement become dependents of the main clause.
These include the fact that the complement 1 controls agreement,
passivizes, hosts PA, and, when raised as a 3, advances to 2 in the union
clause. Evidence that the complement 2 raises comes from the fact that it
is subject to constraints on chomeurs in the union clause, in particular the
constraint which bans reflexive chomeurs, and the person constraint on
chomeurs. The interaction of clause union with the person constraint
(section 2.3.1) is of some interest for the light it sheds on what the person
constraint is not. That constraint rules out 1st and 2nd person chomeurs,
with the consequence that transitive complements in clause union cannot
have 1st or 2nd person 2s (those nominals will always be final chomeurs
in the union clause). It might be imagined that the person constraint is
functionally motivated by a requirement that 1st and 2nd person
arguments be deducible through agreement, together with the fact that
chomeurs cannot control agreement. But in clause union, these arguments
do control agreement on the union predicate (see examples (31)-(32))
making their existence perfectly deducible. Hence, the person constraint
on chomeurs cannot be reduced just to the need for 1st and 2nd person
pronouns to register their existence through agreement.
Plain bivalent stems of the sort discussed in chapter 6, section 2 and this
chapter, section 2.4 occur in a construction which resembles a passive.
Like passive clauses, these appear to contain an intransitive verb - one
systematically related to a transitive verb - and a form of the relational
noun -u7uo, otherwise used to flag passive chomeurs. H Compare (55a)
and (55b). Sentence (55b) is a bona fide passive, while (55a) exemplifies
the construction in question.
(55) a. Lek 7i- poj xa y- u7un i mol Pineda
well cp defend;!, cl A3 u tun the old P.
-e. OCK 173
cl
Old Pineda was able to defend it Ithe town].
A priori, (55a) looks like it should literally translate 'It [the town] was
defended by Old Pineda', with 'it' serving as the 1 of 7ipoj, and yu7un i
mol Pinedae being a passive chomeur. However, at least five facts suggest
to varying degrees that (55a), and similar sentences, are not passives.
First, Tzotzil speakers translate (a)-type sentences into Spanish by
active transitive constructions, while (b)-type sentences are translated by
passives. Second, the (a) sentences have an abilitative sense which is
lacking in the (b) sentences. Third, the intransitive verb in the (a) structure
is not always related to a transitive one, while passive verbs always are.
Jok'i and k'ot are intransitive stems. They cannot be inflected transitively,
and are not related in any productive way to transitive stems. This
effectively eliminates a passive analysis.
Fourth, the only form of -u7un which ever occurs in passive sentences
is yu7un, with a 3rd person possessor (because passive chomeurs cannot
be 1st or 2nd person (see chapter 4, page 63)). But ku7un and avu7un
(with 1st and 2nd person possessors) occur freely in the abilitative
construction:
(58) a. Mu x- ti7 av- u7un. OCK 396
not nt eat;!' A2 u 7un
You can't eat him up.
3.1. Analysis
To say that the clauses in question have VOS order is to claim that they
are superficial simplex clauses. However, a number of facts indicate that
they are not initially monoclausal, but biclausal, with -u7un (roughly 'can
cause') functioning as predicate of the main clause, and taking a clausal
complement whose predicate surfaces as the intransitive verb in this
construction.
232 CHAPTER 11
(63)
-u7un
can cause
-u7un
a
CLAUSE UNIONS 233
The following sections justify various aspects of (64): the initial biclausal
structure, the initial unaccusativity of the complement (which is a
necessary property of this construction), and the final monoclausal
structure.
The most peculiar aspect of the abilitative construction is the surface
order of the two predicates. The general rule in Tzotzil is that the
predicate precedes all other clausal dependents. If -u7un is final predicate
in the union clause, as (64) claims, it should precede the raised verb (the
complement predicate) (just as 7ak' precedes the raised verb in causative
union). This is discussed further in section 3.4.
b. Mu xanav k-u7un.
I can't drive it.
not I can't make him walk.
-u7un
UN
7imane
be bought
Sentence (73) would instantiate (72), but (73) does not have an abilitative
reading. It is unambiguously passive.
(73) 7i- man -e y- u7un Ii Xun -e.
cp buy psv A3 u7un the Xun cl
It was bought by Xun.
not: Xun was able to buy it.
Cf. 7iman yu7un Ii Xune 'Xun was able to buy if, where 7iman is the
intransitive form of the bivalent stem man.
Further, if passive complements were possible, then combining a
passive verb with ku7un or avu7un should yield an abilitative with a 1st
or 2nd person 1. But it yields instead a passive clause with an oblique
phrase interpreted as cause (ku7un and avu7un cannot be construed as
agent phrases because they are not 3rd person).
(74) 7i- maj -e k- u7un/ av- u7un Ii Xun -e.
cp hit psv Al u7un/ A2 u7un the Xun cl
Xun was hit because of me/you.
236 CHAPTER 11
a -u7un b
can cause
(I will in fact argue that the surface structure is something like this.) The
crucial difference between (64) and (75) is that in the former, the final 2
in the main clause is also final 2 in the complement clause. The union
analysis represented by (64) predicts certain agreement facts, namely that
the complement predicate agrees with its final 2. Under (75), a never itself
has any syntactic argument. Nothing yet said makes any predictions about
the inflection on this verb. However, the nominal which corresponds to b
in (75) must be cross-referenced on a. This follows immediately under the
biclausal analysis represented in (64): b is final 2 of the clause whose
predicate is a. The agreement rules of chapter 3 require that a agree in
person with b.
CLAUSE UNIONS 237
The agreement rules of chapter 3 require that the final dependents of the
union clause be cross-referenced on the predicate of that clause. Take
(76), for example.
238 CHAPTER 11
(82)
av-u7un
A2-can cause
x-i-toj
nt-B}-pay;v
As final ergative in the union clause, the 2nd person pronoun must be
cross-referenced on the predicate by an A2 prefix: this is apparently
sa!isfied by the prefix on avu7un. The 1st person pronoun is final
absolutive in the union clause and must be cross-referenced on the
predicate by a B1 affix. The only B1 affix in (76) is attached to xitoj. The
fact that the relevant agreement affixes are shared between xitoj and
avu7un, suggests that the two together constitute the final predicate in the
union clause. Under the usual analysis of union constructions, the comple-
ment predicate is not (part of) the union clause predicate. It is a
dependent in the union clause, but bears the U relation. I assume this, but
assume in addition that the predicate and union verbs combine in a
subsequent stratum to form a new predicate, as sketched below in (83).
(83)
c
CLAUSE UNIONS 239
At a fairly gross level, the structure proposed in (83) makes the right
predictions about agreement. But one can do better than simply point out
that affixes of the appropriate type appear somewhere in this compound
predicate. Structure (83), together with the agreement rules of chapter 3,
makes a set of fine-grained predictions about where in this compound
predicate the various affixes can appear. These are summarized below,
and refer to elements a, b, c, d, e in (83).
(A) The final absolutive in the complement clause (i.e., a) must be
cross-referenced on the complement predicate, c (because agreement with
final 1 and 2 is obligatory).
(B) Only the final absolutive in the complement clause (i.e., a) can be
cross-referenced on c (because any agreement marker on the complement
predicate must cross-reference a final dependent in the complement
clause).
(C) The final absolutive and ergative in the union clause (i.e., a and b)
and only these must be cross-referenced on the union predicate, e (same
reasoning as above).
(A) and (C) together require that a be cross-referenced both on the
240 CHAPTER 11
The prefix -i- and the suffix -otik both cross-reference a (a = 1st person
plural inclusive pronoun). The prefix -i- must appear on xitoj (c) because
person agreement is obligatory in the complement clause. This prefix
simultaneously satisfies person agreement rules in the union clause. The
suffix -otik is not required on xitoj because number agreement is optional.
It may, however, appear on -u7un because the 1st person pronoun (a) is
final absolutive in the union clause as well, and -u7un is part of the union
predicate, e.
In (85) the 2nd person pronoun is cross-referenced by both the B2
prefix -a- and the plural suffix -ik:
The prefix -a- must appear on xatoj (c) to satisfy person agreement within
the complement clause. At the same time, its appearance on xatoj satisfies
person agreement rules in the union clause, since xatoj is part of the union
predicate. Again, -ik is not required on xatoj because number agreement
is optional. On the other hand, it can appear on d (-u7un) because the
2nd person pronoun is final absolutive in the union clause, and d is part
of the union predicate, e.
As expected, the plural affixes can affix to e instead:
ku7unbe
can cause
x kuch
carry;v
smoch he
basket
4. SUMMARY
32A, -be attaches to 7ak'. The verb of the complement clause has none
of these features - facts which follow if it is assigned the Union relation in
the union clause. In contrast, the relation of the two predicates in
abilitative union is more complex. The distribution of agreement affixes
supports the view that the intransitive verb in the construction is the final
predicate of the complement clause, that -u7un is the initial predicate of
the main clause, and that the two combine to form the final predicate of
the union clause.
The following section contains a brief discussion of relevant APG laws
and rules. It sketches the APG approach to union constructions and the
required Tzotzil rules.
NOTES
I See also Davies and Rosen (1985) for a very interesting monoclausal analysis of clause
union within RG. See Postal (1986b) for a recent APG treatment in which all complement
nominals raise as Dead, and then in some cases advance to other relations, determined in
part by factors like those discussed here.
2 Example (i), which does not involve P A, has the coreferential reading:
This is because all the examples are imperatives, selected to provide very clear contrasts
between unreduced and union causatives. If the complement 2 were 2nd person, there
would be coreference between the main clause 1 and the complement 2, a configuration
blocked independently (see next section).
5 Laughlin's translations of (44) and (46) are 'He entered to let Itheml see his money', and
'That woman let lhiml see it', brackets mine. The pronouns in brackets are understood, but
appear to correspond to no syntactic element.
6 Rosen (1983) argues, on the basis of stronger evidence than that presented here, for
such a constraint in Italian, and suggests it holds more generally in Romance. Chun et al.
(1984) and Gerdts (1986) argue that no such constraint holds in Korean, concluding that
its presence in grammars must be guaranteed by language-particular rules.
248 CHAPTER 11
7 Structure (50) violates the Final 1 Law. At worst, the complement in union constructions
is an exception to this law. At best, this exceptionality would follow from something, and
not need to be stipulated. Davies and Rosen (1985) cite this as a consequence of their
account.
S Indeed, Cowan calls these "pseudo-passive transitive clauses" (1969, p. 48). She goes on
to say, "[the constructionl is grammatically anomalous in two respects. First, it has a goal as
subject and the agent constituent is based on the associative noun stem u7un as in a true
transitive passive clause. Second, the verb stem - which may be intransitive or transitive -
occurs with active intransitive inflection." I argue below that, despite appearances, these
clauses do not have a goal as subject, and the verb stem is always intransitive. Haviland
(1981, pp. 276-9, 292) mentions this construction, and appears to regard instances of it
as passive transforms from active transitive sentences: "they are transformed into
mediopassives by means of -u7un" [translation minel.
4 Cowan (1969) claims that transitive verb stems can occur in this construction, probably
referring to stems like poj, ti7, jam, k'el. My claim is that these stems can be used either
transitively or intransitively, but that here, they are intransitive. The appearance of
transitive-looking stems in abilitatives is due simply to the fact that monosyllabic transitive
stems have identical intransitive counterparts. Polysyllabic transitive stems cannot occur in
this construction. Consider, for example, the pair jitun, jituj. Jitun is a transitive verb stem,
meaning 'untie'; jituj is an intransitive stem, meaning 'become untied'. The same semantic
relation that holds between transitive and intransitive versions of bivalent stems holds here,
and is mediated by the alternation -un/-uj. (-Vn/-Vj, where the vowels are identical, is a
fairly common means for relating transitive and intransitive verb pairs. Cf. k'opon 'speak
to, address', k'opoj 'speak'; 7elk'an 'steal', 7elk'aj 'steal'; meltzan 'make', meltzaj 'be
made'.) Only the intransitive verb is possible in this construction:
(i) *Mu x- jitun k- u7un Ii j- ka7 -e.
not nt untie Ai u7un the Ai horse c/
(I can't untie my horse.)
believe it cannot, but my data is not clear enough on this point to support a position. It is
also unclear whether 3s other than ascended possessors occur in this construction. I have
elicited a few examples involving 3s interpreted as recipients or benefactives, but have seen
no textual examples of this type and am not convinced they exist.
II In so-called Equi unions of the type discussed in Aissen and Perlmutter (1983), the
erasure of one final arc is motivated independently of the union structure, namely, that
CLA USE UNIONS 249
final arc whose head is the Equi ·victim'. Hence, this arc has no foreign successor in the
union clause. In general, only those final arcs whose erasure is not motivated indepen-
dently of the union structure have foreign successors in the union clause. See J & P,
chapter 8.6.
12 Again, the U arc locally sponsors all those foreign successors whose existence is not
motivated independently of the union construction. In the case of Raising unions, discussed
also in Aissen and Perlmutter (1983), the arc headed by the ascendee is sponsored by its
support, as is required in ascensions, and not by the U arc. See J & P, chapter 8.6.
IJ Law 101 differs from analogous laws in J & P and in Postal (1986b).
14 Cho arc launchers raise certain problems. Under (101), a Cho arc launcher has a Cho
arc foreign successor, violating the Chomeur Law, which excludes Cho arc foreign
successors. My solution at present would be to modify the latter to allow Cho arc foreign
successors only in the case of Cho arc launchers.
APPENDIX TO CHAPTER 11
(8) Ch- av- ak' -be s- ti7 choy ... OCK 315
icp A2 let io A3 eat fish
You let them eat fish ...
250 CHAPTER II
(12) Mu yechuk n070x ch- k- ak' av- uch' -ik Ii v07 -e. OCK 146
not thus just icp Al let A2 drink pI the water el
I'm not going to let you drink the water for nothing.
(14) 7i- y- ak' 7i- p'is -b -at s- tak'in Ii jun vinik -e. OCK 150
cp A3 let cp count io psv A3 money the a man el
The man let his money be counted.
Ambiguous between Unreduced Causative and Union Causative
(IS) 7ak' -0 7abtej -uk. OCK 36
let imp work subj
Have him work!
(24) 7aver buch'u x- ch'oj y- u7un i te7 ta k'obol -e. OCK 383
to see who nt perforate,v A3 u tun the tree with fist cl
To see who can make a hole in the tree with his fist ...
Other Intransitive Stems in Complement
(25) Ti mi balch'uj av- u7un -e ... OCK 81
if roll A2 u7un cl
If you can roll them ...
(29) Pero mu7 nux x- k'ot y- u7un i kareta -e. OCK 215
but not ever nt arrive A3 u tun the cart c/
But a cart can't ever get it there.
1. INTRODUCTION
2. QUANTIFIERS
a b
256 CHAPTER 12
a b
QUANTIFICATION AND INITIAL ABSOLUTIVES 257
Here again a is the initial absolutive, and b the final absolutive. Pre-
predicate 7ep binds the initial absolutive:
But under the present analysis of reflexive clauses, the final ergative is the
initial absolutive, for 7amuchetike 'toads' heads both the initial 1 and 2
arcs, with the reflexive nominal sba entering in a subsequent stratum.
According, 7amuchetike is the initial absolutive, yielding exactly the right
prediction about the interpretation of prepredicate quantifiers in reflexive
clauses. This provides very strong evidence for the RG/ APG representa-
tion of coreference involving initial overlapping arcs, for under that
account it is entirely coherent to assert that the final ergative is the initial
absolutive. Closely related facts provide the basis for rejecting a surfacist
account of quantifier binding in Tzotzil (see section 6, below).
Furthermore, the apparent binding of the ergative in (20)-(22) is not a
general property of reflexive clauses for in ditransitive reflexive clauses,
prepredicate 7ep does not bind the ergative:
258 CHAPTER 12
This too follows directly from the analysis of ditransitive reflexive clauses,
for the ergative does not head an initial absolutive arc (see chapter 7,
section 3.3). The initial absolutive is kalto in (23) and yosH in (24). As
predicted, 7ep binds the initial absolutive. Hence, the apparent binding of
the ergative in examples like (20)-(22) is due simply to the fact that in
monotransitive reflexive clauses, the final ergative and initial absolutive
overlap.
The relevant feature of these examples is that the initial absolutive cannot
QUANTIFICATION AND INITIAL ABSOLUTIVES 259
But even here there is a clear preference for binding the initial
absolutive. When both the initial absolutive and final absolutive satisfy the
classifier, only the initial absolutive is bound:
In general, then, it appears that if the quantifier can bind the initial
absolutive, it will. If it cannot, then and only then may it bind the initial
3/final 2. Speaker reaction to (26) is instructive. One interpretation is
indicated in the translation. But one informant, laughing, translated it, 'I
gave three pieces of candy to the boys'. He explained that the candies had
to be shaped like humans. Noun stems are not arbitrarily assigned to
classifier sets; rather a classifier imputes properties to the referent of the
relevant nominal. If an interpretation is available under which the initial
absolutive is bound, literal or otherwise, that is the preferred interpreta-
tion. Of course, speakers will vary in the interpretations they 'see', both
among themselves and from occasion to occasion.
A similar but much weaker effect is observed with ergatives. Where
both the ergative and initial absolutive satisfy the classifier, only the
absolutive is bound:
260 CHAPTER 12
But when the ergative is the only nominal which satisfies the classifier,
then speakers occasionally accept such sentences, while judging them
marginal.
The data discussed above follow from two principles. The first is that
only absolutives are bound:
By (35), no nominal that heads only an ergative arc can be bound. This
explains the absence of the ergative-binding reading in (8)-(10), and the
marginality of examples like (33)-(34). Furthermore, it allows for ergative
binding in mono transitive reflexive clauses because the ergative is also
initial absolutive.
The second principle will allow a nominal which does not head an
initial absolutive arc to be bound just in case the initial absolutive cannot
be. In other words, if both the initial and final absolutive are eligible for
binding, the initial absolutive takes precedence. This transderivational
condition does not constrain representations of individual sentences, but
rather quantifies over sets of such representations. (36) must be viewed as
extremely informal since the last condition ("C corresponds to d") remains
undefined.
QUANTIFICATION AND INITIAL ABSOLUTIVES 261
5. POSTPREDICATE QUANTIFIERS
7ep itself is not a nominal dependent, but 7epal, a related word, is (d.
chapter 1, section 3.3). Both (41) and (42) provide evidence for this.
Statement (55) accounts for the fact that prepredicate 7ep binds only the
initial absolutive, because the initial absolutive is (almost) always the first
nominal to its right. Prepredicate cardinal numbers can bind the initial
3/final absolutive, passing over the initial absolutive, just in case binding
the latter results in semantic ill-formedness. In postpredicate position, 7ep
binds the nominal which immediately follows it, either the final chomeur
or the final 2.
(though (36) may require modification for both languages) - rules which
refer to grammatical relations and not to dominance or linear order.
Tzotzil also has (46), which requires that a quantifier precede the nominal
it binds. Pima lacks (46).
This is not to say however that Pima has no constraints on the relative
positions of quantifier and bound nominal. Munro notes that there is a
strong tendency for a nominal to follow a quantifier that binds it (Le., a
tendency towards (46», and she observes (1984, fn. 12) that word order is
more rigid than otherwise in sentences involving quantifier binding. All the
examples Munro cites where a quantifier follows the bound nominal have
the following character:
Subject - Auxiliary - Q - X
7. CONCLUSION
NOTES
complex nominal:
(i) 7ep kremotik ba s- man -ik tal kaxlan vaj ta ch'ivit.
lots boys went A3 buy pi coming bread in market
Many boys went to buy bread in the market.
(lit: Many are the boys who went to buy . .. )
APPENDIX TO CHAPTER 12
Prepredicate Quantifiers
(1) 7ep s- tzak jay -v07 y- aj7ikatznom ti solteroetike. OCK 60
many A3 grab how many nc A3 porters the soldiers
The soldiers seized many porters.
270 CHAPTER 12
(9) 7ep ch- [yj- il -ik tajmek jk'ux-7ak'al tajmek. OCK 320
lots icp A3 see pi lots charcoal-cruncher lots
They saw a lot of charcoal crunchers.
(10) Mu7 no j- na7 k'u yu7un ti toj tol 7ep 7i- k- uch'
not cI Al know what for the very much much cp Al drink
i v07 -e. OCK 285
the water cI
I don't know why I drank so much water.
(12) Cha7- 10m j- lap -oj Ii j- natil vex -e. SSS 113
2 nc A I wear pf the Ai long pants cI
I was wearing two pairs of long pants.
(19) 7i- 7ak' -b -at la 7ep bek'et, vaj, kaxlan vaj. OCK 74
cp give io psv cl much meat tortilla bread
It was given a lot of meat, tortillas, and bread.
(20) 7i- lok'es -b -at tal 7ep Ii k'u7uletik -e. SSS 129
cp remove io psv here much the clothes cl
He took out lots of clothes for them.
integrated work must aim for theoretical consistency, and while the task is
arduous, the gain in analytical power is significant. By imposing a
consistent set of assumptions, it becomes possible to draw conclusions
across diverse parts of an analysis. Generalizations about the analysis itself
can emerge, and it is these which constitute the theoretically significant
results of any study.
Consider two examples from the present work. Each represents a
generalization about the analysis itself, and each draws from diverse pieces
of the analysis.
The first concerns the case for a multistratal syntax. A number of the
rules proposed express generalizations across superficially diverse linguis-
tic elements. The essential claim of a multistratal theory is precisely that
such generalizations exist, and the strongest argument for a multistratal
theory would be that such generalizations cannot be represented at all in
monostratal accounts. Whether they can or not remains to be seen, but
there is no doubt that cases like the following constitute an interesting
challenge for monostratal theories.
(i) The principles of quantifier binding (chapter 12) generalize across
the class of initial absolutives. On the one hand, this class includes surface
ergatives in reflexive clauses. On the other, it includes surface chomeurs in
ditransitive clauses and surface absolutives in mono transitive and intransi-
tive clauses.
(2) The account of surrogate agreement (chapter 10) involves a
generalization over ascension hosts: possessor ascension hosts and
conjunct union hosts. Superficially, these elements are entirely distinct: PA
hosts are possessed nominals while conjunct union hosts never occur in
surface structure at all. Under the ascension analysis, they are both
ascension hosts.
(3) The person condition on advancement chomeurs (chapter 7) is a
generalization across the objects of certain PPs in passive clauses, the
possessors of certain nouns in others, and certain unflagged nominals
in ditransitive clauses. Under advancement analyses for passives and
ditransitive clauses, these elements have in common that they are
chomeurs.
(4) The coreference condition on certain 'last' 2s in monotransltIve
clauses (chapter 9) involves a generalization across certain surface 2s,
certain surface 1s of passive clauses, and certain topicalized elements.
Under the proposed analyses, each of these elements is the 'last' 2 in its
clause.
(5) The flagging rule for conjuncts (chapter 10) applies to entirely
distinct surface elements: to surface conjuncts, which are always depen-
dents of a coordinate node, and to surface elements which are clausal
dependents. Under the conjunct union account, the latter are earlier
conjuncts and the flagging rule can apply simply to conjuncts.
274 CONCLUSION
Zinacantec Tzotzil has the following consonantal phonemes, represented below in the
orthography used in this study.
Glottalized
V-less Stop p f k'
V-d Stop b
V-less Affricate tz' ch'
Spirant
V-less (f) s x j
V-d (w) v Y
Liquid r, I
Nasal m n
Sounds which occur only in recent loans are enclosed in parentheses. In addition, Tzotzil
has five vowel phonemes, represented by i, e, a, 0, and u. See also Colby (1964), Hopkins
(1962, 1967), and Weathers (1947).
A number of morpho phonological rules alter the shape of stems and inflectional affixes.
For present purposes, the set of underlying segments can be identified with the set of
phonemes. In the rules which follow, underlying segments too are represented by the
conventional orthography.
Stem-initial glottal stop deletes after set A prefixes. The glottal stop appears in morpho-
logically related forms not bearing set A prefixes. For example, the active forms of
transitive verb stems with initial glottal stop lose the stop, while passive forms, which lack
set A prefixes, retain it.
(I) 7i-y-7ak'-be (cp-A3-give-io) - 7iyak'be 'he gave it to him'
(d. 7i7ak'bat (cp-give-io-psv) 'he was given if)
7i-k-7il (cp-A I-see) - 7ikil 'I saw it'
(d. 7i7ilat (cp-see-psv) 'it was seen')
275
276 PHONOLOGICAL RULES
Initial glottal stop drops in the possessed form of [7J-initial nouns, but it appears in
other derived forms.
(2) 7av-7ol (A2-child) -+ 7avol 'your child'
(cf.j7alnom (agn-child-nominalizer) 'one who has given birth')
k-7ajnil (Al-wife) -+ kajnil 'my wife'
(cf. j7ajnilajel (agn-wife-nominalizer) 'groom')
The prevocalic variant of the 3rd person set A prefix, y-, drops after the neutral aspect
prefIX X-. (X- marks not only neutral aspect, but in combination with la, marks incomple-
tive aspect.)
(3) ta x-y-il (icp-A3-see) -+ ta xii 'he'll see it'
(cf. 7iyil (cp-A3-see) 'he saw it')
mu x-y-ik' (not nt-A3-marry) -+ mu xik' 'he's not marrying'
(cf. yik'oj (A3-marry-pf) 'he has married her')
The neutral aspect marker x- drops before those set A prefixes which are spirants: s- (A3,
preconsonantal) and j- (A 1, preconsonantal).
(4) ta x-s-muk (icp-A3-bury) -+ ta smuk 'he buries it'
(cf. ta x-a-muk (icp-A2-bury) -+ ta xamuk 'you'll bury it')
ta x-j-man (icp-Al-buy) -+ tajman 'I'll buy it'
(cf. ta x-a-man (icp-A2-buy) -+ ta xaman 'you'll buy it')
Note that x- appears before all prevocalic set A prefixes.
(5) ta x-k-il (icp-Al-see) -+ ta xkil 'I'll see it'
ta x-y-il (icp-A3-see) -+ ta xii 'he'll see it' (see section 2)
This suggests that the deletion of x- is phonologically conditioned. X- fails to drop before
01, however, when UI is a stem-initial segment. (Before stem-initial [sJ, x- generally
assimilates to [s] by Spirant Assimilation (section 4), and the resulting geminate cluster is
simplified by Geminate Reduction (section 6).)
(6) ta x-jatav (icp-flee) -+ ta xjatav 'he'll flee'
ta x-jam (icp-open) -+ ta xjam 'it'll open'
In summary, the neutral aspect prefix x- deletes before those set A prefixes which are
spirants.
4. SPIRANT ASSIMILATION
5. CONTRACT I ON
The sequences' Ial Is· and' Ial h· optionally con tract to 'U· and 'ch· res~cti vely,
TI has two functions; it is a preposition, and compounded with verbs in neu tral aspect, it
form s incompletive aspect. It undergoes con traction in both functions,
'Tal Is· has Ihree sources: (i) the preposition II plus a noun plefixed with s- (A3,
preconsonantal): (il) verb in incompletive aspect prefixed with s' (A3, pleconsonanlal): !I
x-s-Verb - t.a s-Verb (sec section 3 for deletion of x-J: (iii) sequences of II s- which arise
by spi rant assimilation from II x- (see seclion 4).
6. GEMINATE REDUCTION
The following clusters are simplified: 1771, liiI.lnnl, Issl.lxxJ (Laughlin 1975, p, 26).
278 PHONOLOGICAL RULES
7. VOWEL DELETION
A number of inflectional suffixes are subject to reduction by vowel syncope. These include
the following cases.
7.2. -be
The predicate suffix -be, associated with indirect objects, reduces to -b before -a and -0.
Aissen, Judith: 1979, 'Possessor Ascension in Tzotzil', in Laura Martin, ed., Papers in
Mayan Linguistics, Lucas Brothers, Columbia, Missouri, pp. 89-108.
- - - - : 1982, 'Valence and Coreference', in Sandra Thompson and Paul Hopper, eds.,
Syntax and Semantics, Vol. 15, Academic Press, New York, pp. 7-35.
- - - - : 1983, 'Indirect Object Advancement in Tzotzil', in David M. Perlmutter, ed.,
Studies in Relational Grammar J, University of Chicago Press, Chicago, pp. 272-302.
- - - - : 1984a, 'Themes and Absolutives: Some Semantic Rules in Tzotzil', in Donna
Gerdts and Eung-Do Cook, eds., Syntax and Semantics, Vol. 16, Academic Press, New
York,pp.I-20.
- - - - : 1984b, 'Control and Command in Tzotzil Purpose Clauses', Proceedings of the
Tenth Annual Meeting of the Berkeley Linguistics Society, University of California,
Berkeley, pp. 559-71.
- - - - : to appear, 'Extensions of Brother-in-law Agreement', Agreement in Natural
• Language: Approaches, Theories, and Descriptions, Center for the Study of Language
and Information, Palo Alto.
Aissen, Judith and David M. Perlmutter: 1983, 'Clause Reduction in Spanish', in David M.
Perlmutter, ed., Studies in Relational Grammar J, University of Chicago Press,
Chicago, pp. 360-403.
Allen, Barbara J., Donna B. Gardiner, and Donald G. Frantz: 1984, 'Noun Incorporation
in Southern Tiwa', Internationallournal of American Linguistics 50, 292-311.
Ayres, Glenn: 1980, 'A Note on Mayan Reflexives', Journal of Mayan Linguistics 2, 53-9.
Bell, Sarah: 1983, 'Advancements and Ascensions in Cebuano', in David M. Perlmutter,
ed., Studies in Relational Grammar J, University of Chicago Press, Chicago, pp. 143-
218.
Berlin, Brent: 1968, Tzeltal Numeral Classifiers: A Study in Ethnographic Semantics,
Mouton, The Hague.
Bresnan, Joan: 1982, ed., The Mental Representation of Grammatical Relations, MIT
Press, Cambridge, Massachusetts.
Bricker, Victoria: 1977, Pronominallnj1ection in the Mayan Languages, Middle American
Research Institute, Tulane University, New Orleans.
Chun, Soon Ae, Donna Gerdts, and Cheong Youn: 1984, 'Unaccusatives and Clause
Union in Korean: The Downstairs Thaw', paper presented at the Symposium on
Grammatical Relations, State University of New York, Buffalo.
Colby, Lore: 1964, Zinacantan Tsotsil Sound and Word Structure, unpublished Ph.D.
dissertation, Harvard University.
Cole, Peter and S. N. Sridhar: 1977, 'Clause Union and Relational Grammar: Evidence
from Hebrew and Kannada', Linguistic Inquiry 8,700-13.
Cowan, Marion: 1969, Tzotzil Grammar, Summer Institute of Linguistics, Norman,
Oklahoma.
Cowan, Marion and Wm. Merrifield: 1968, 'The Verb Phrase in Huixtec Tzotzil',
Language 44, 284-305.
Davies, William D. and Carol Rosen: 1985, 'Unions as Multi-Predicate Clauses',
unpublished manuscript, Department of Modern Languages and Literatures, Cornell
University, Ithaca, New York.
Delgaty, Colin: 1960, 'Tz~tzil Verb Phrase Structure'. Mayan Studies I, Summer Institute
of Linguistics, Norman, Oklahoma, pp. 81-125.
279
280 REFERENCES
Dixon, R. M. W.: 1972, The Dyirbal Language of North Queensland, Cambridge University
Press, Cambridge.
Fauconnier, Giles: 1983, 'Generalized Union', Communication and Cognition 16,3-37.
Fleck, Margaret M.: 1981, 'Tzotzil Numeral Root Morphology', Journal of Mayan
Linguistics 3, 5-24.
Frantz, Donald G.: 1979, 'Multiple Dependency in Blackfoot', Proceedings of the Fifth
Annual Meeting of the Berkeley Linguistics Society, University of California, Berkeley,
pp.77-80.
Gazdar, Gerald, Ewan Klein, Geoffrey Pullum, and Ivan Sag: 1985, Generalized Phrase
Structure Grammar, Basil Blackwell, Oxford.
Gerdts, Donna: 1986, 'Causatives and Passives in Korean: Evidence for Clause Union
Without Revaluation', unpublished manuscript, Department of Linguistics, State
University of New York, Buffalo.
Gibson, Jeanne and Eduardo Raposo: 1986, 'Clause Union, the Stratal Uniqueness Law,
and the Chomeur Relation', Natural Language and Linguistic Theory 4, 295-33l.
Hale, Kenneth: 1975, 'Counterexamples and Explanations in Navajo Linguistics: Syntax',
Navajo Language Review 2, 29-60, Center for Applied Linguistics, Washington, D.C.
Hankamer, Jorge and Ivan Sag: 1976, 'Deep and Surface Anaphora', Linguistic Inquiry 7,
391-426.
Harris, Alice c.: 1981, Georgian Syntax: A Study in Relational Grammar, Cambridge
University Press, Cambridge.
- - - - : 1984, 'Inversion as a Rule of Universal Grammar: Georgian Evidence', in
David M. Perlmutter and Carol Rosen, eds., Studies in Relational Grammar 2, Uni-
versity of Chicago Press, Chicago, pp. 259-9l.
Haviland, John: 1981, Sk'op Sotz'leb: El Tzotzil de San Lorenzo Zinacantan, Universidad
Nacional Aut6noma de Mexico, Mexico City.
Hopkins, Nicholas: 1962, A Phonology of Zinacantan Tzotzil, unpublished MA thesis,
University of Texas.
- - - - : 1967, 'A Short Sketch of Cha1chihuitan Tzotzil', Anthropological Linguistics 9,
9-25.
Johnson, David E. and Paul M. Postal: 1980, Arc Pair Grammar, Princeton University
Press, Princeton, New Jersey.
Joseph, Brian: 1976, 'Raising in Modern Greek: A Copying Process?', in Jorge Hankamer
and Judith Aissen, eds .. Harvard Studies in Syntax and Semantics, Vol. 2, Harvard
University, Cambridge, Massachusetts, pp. 241-278.
- - - - : 1978, Morphology and Universals in Syntactic Change: Evidence from Medieval
and Modern Greek, unpublished Ph.D. dissertation, Harvard University.
Kaufman, Terrence: 1971, Tzeltal Phonology and Morphology, University of California
Press, Berkeley.
Kayne, Richard: 1975, French Syntax, MIT Press, Cambridge, Massachusetts.
Keenan, Edward L.: 1974, 'The Functional Principle: Generalizing the Notion of "Subject
of''', Papers from the Tenth Regional Meeting of the Chicago Linguistic Society,
Chicago, pp. 298-309.
Kisseberth, Charles and Mohammad Imam Abasheikh: 1977, 'The Object Relationship in
Chi-Mwi:ni, A Bantu Language', in P. Cole and J. Sadock, eds., Syntax and Semantics,
Vol. 8, Academic Press, New York, pp. 179-218.
Lakoff, George and Stanley Peters: 1969, 'Phrasal Conjunction and Symmetric Predicates',
in D. A. Reibel and S. A. Schane, eds., Modern Studies in English, Prentice-Hall,
Englewood Cliffs, N.J., pp. 113-42.
Laughlin, Robert: 1975, The Great Tzotzil Dictionary of San Lorenzo Zinacantan,
Smithsonian Institution Press, Washington, D.C.
- - - - : 1977, Of Cabbages and Kings, Smithsonian Institution Press, Washington, D.C.
REFERENCES 281
- - - - : 1980, Of Shoes and Ships and Sealing Wax, Smithsonian Institution Press,
Washington, D.C.
Munro, Pamela: 1984. 'Floating Quantifiers in Pima', in E.-D. Cook and D. B. Gerdts, eds.,
Syntax and Semantics, Vol. 16, Academic Press, New York, pp. 269-87.
Norman, William: 1978, 'Advancement Rules and Syntactic Change: The Loss of Instru-
mental Voice in Mayan', Proceedings of the Fourth Annual Meeting of the Berkeley
Linguistics Society, University of California, Berkeley, pp. 258-76 ..
Partee, Barbara Hall: 1971, 'On the Requirement that Transformations Preserve Meaning',
in Charles 1. Fillmore and D. Terence Langendoen, eds., Studies in Linguistic
Semantics, Holt, Rinehart and Winston, Inc., New York, pp. 1-21.
Perlmutter, David M.: 1978, 'Impersonal Passives and the Unaccusative Hypothesis',
Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society, Univer-
sity of California, Berkeley, pp. 157-89.
- - - - : 1983, 'Personal vs. Impersonal Constructions', Natural Language and Linguis-
tic Theory 1, 141-200.
- - - - : 1984, 'Working Is and Inversion in Italian, Japanese, and Quechua', in David
M. Perlmutter, ed., Studies in Relational Grammar 2, University of Chicago Press,
Chicago, pp. 30-80.
Perlmutter, David M. and Paul M. Postal: 1974, Lectures on Relational Grammar, Summer
Linguistic Institute of the L.S.A., University of Massachusetts, Amherst.
- - - - : 1983a, 'Some Proposed Laws of Basic Clause Structure', in David M.
Perlmutter, ed., Studies in Relational Grammar I, University of Chicago Press,
Chicago, pp. 81-128.
- - - - : 1983b, Toward a Universal Characterization of Passivization', in David M.
Perlmutter, ed., Studies in Relational Grammar I, University of Chicago Press,
Chicago, pp. 3-29.
- - - - : 1983c, The Relational Succession Law', in David M. Perlmutter, ed., Studies in
Relational Grammar I, University of Chicago Press, Chicago, pp. 30-80.
- - - - : 1984, 'Impersonal Passives and Some Relational Laws', in David M. Perlmutter
and Carol G. Rosen, eds., Studies in Relational Grammar 2, University of Chicago
Press, Chicago, pp. 126-170.
Perlmutter, David M. and Scott Soames: 1979, Syntactic Argumentation and the Structure
of English, University of California Press, Berkeley.
Postal, Paul M.: 1986a, Studies of Passive Clauses, State University of New York Press,
Albany, New York.
- - - - : 1986b, 'Masked Inversion in French Clause Union Complements', unpublished
paper, IBM, Yorktown Heights, New York.
Rosen, Carol: 1981, The Relational Structure of Reflexive Clauses: Evidence from Italian,
unpublished Ph.D. dissertation, Harvard University.
- - - - : 1983, 'Universals of Causative Union: a Co-Proposal to the Gibson-Raposo
Typology', Papers from the 19th Regional Meeting of the Chicago Linguistic Society, pp.
338-52.
- - - - : 1984, The Interface between Semantic Roles and Initial Grammatical
Relations', in David M. Perlmutter, ed., Studies in Relational Grammar 2, University of
Chicago Press, Chicago, pp. 38-77.
Ross, John R.: 1967, Constraints on Variables in Syntax, Ph.D. dissertation, MIT,
published by the University of Indiana Linguistics Club, Bloomington, Indiana (1968).
Weathers, Nadine: 1947, Tzotzil Phonemes with Special Reference to Allophones of b',
International JQurnal of American Linguistics 13, 108-111.
- - - - : 1950, 'Morphological Analysis of a Tzotzil (Mayan) Text', International
Journal of American Linguistics 16, 91-8.
Williams, Edwin: 1978, 'Across-the-Board Rule Application', Linguistic Inquiry 9, 31-43.
INDEX
Note: Words beginning with 7 precede all other entries; 3 and 32A follow all other entries.
Pages in italics contain definitions or explanations of terms, rule statements, etc. Page
numbers followed by app contain relevant examples from appendices.
7a, 17-8, 157. See also Topicalization genitive, 4, 19n3, 40, 43-4, 46, 48, 49,
7ak', 15,214,244.246. See also Causative 5~ 5~60nll, 78,8~ 81,83, 127-
clause union 8, 152n2, 178n3
70, 8, 9, 14, 93 in abilitative clause union, 237-242,
70y,6-7 244,250-1 app
A stems, 2, 5-7, 19n3 in causative clause union, 215- 7,
Abasheikh, M., 119 249app
Abilitative clause union, 213, 229-247, in conjunct union, 183, 188-9, 194-5,
248n8, 248n9, 248n 10, 250app, 210-lapp
251app in ditransitive clauses, 107-109. 117,
agreement, 236-244 121, 124-5app, 180-1
coreference conditions, 243 in passive clauses, 61, 62, 64-65, 74n3,
evidence for biclausal structure, 231, 181-2
236,237 in possessor ascension, 127-131, 132,
meaning, 232 151, 154app,210app
possessor ascension, 242-244 in purpose clauses, 16-17
restrictions on complement, 233-235 in reflexive clauses, 78,80,81,83, 86n4
Tzotzil Abilitative Union Rule, 247 number, 40, 46-53. See also Surrogate
Tzotzil Ohligatory Union Trigger Rule, agreement
246 person, 40, 43-46
word order, 231,233,239 supports, 55, 57, 59nl1,60nl1
Ahsolutive, 2 Tzotzil rules, 57-58
Abs(olutive) arc, 25 Aissen, 1., 16, 37, 152n3, 177n3, 204,
agreement, 40, 43, 47-49, 57-8, 59n3, 209n10,210nI3,248n11,249nI2
62,64, 74n3, 107, 127, 131, J81- Allen, B.. 147, 153n8
2, 203, 215-217, 219, 237-241. Anaphoric
See also Set B affix arc, 79, 82, 89, 90, 99, 111, 113, 123,
initial absolutive, 252, 254-261, 263- 133-4,161,174,191,210nI5
268 pronoun, 29, 79-80, 86n7, 112-3, 134,
Accusative, 136, 138, 148 150,168,228
Achenese,210nl3 Anaphorically connected, 79, 81, 82, 85,
Adverbials, order of, 18 89, 100, 111, 122, 123, 134, 150,
Affix (Af) relation, 54, 73, 119 191, 192, 203
Agent phrase, 93, 226. See also Passive Ancestral relations, 31. See also R(emote)-
chomeur Animacy/person hierarchy, 40. See also
Agentive nouns, 19n 1 Conjunct Union, person hierarchy
Agreement, 1, 40ff, 213, 217, 252. See also Antecede, 79, 85, 161, 162, 163, 167, 173,
Surrogate agreement 174. See also Reflex-antecede; Recip-
affixes, 54-5 antecede
APG account, 26, 54-58 Anticopy, 177
arc, 55 arc, 161, 174, 175, 179n 7
ergativity of, 2 pronoun, 155,161, 162, 163
282
INDEX 283