by Maura Velazquez

Department of Linguistics, UCSD

1. Introduction

     In this paper, I will discuss the  different  ways
in  which  the  concept  of  Possession  is  encoded in
Guarani. I will characterize the  conceptual  basis  of
each  of  the  different  possessive constructions, and
will argue that the formal distinctions reflect  under-
lying  semantic  differences.  I  will  show  that even
though Guarani does not have special morphology to dis-
tinguish  alienable  and inalienable nouns, such a dis-
tinction plays a crucial role in explaining the  formal
differences between possessive constructions. The paper
will    also    show    that     apart     from     the
alienable/inalienable  distinction, there are finer se-
mantic differences (such as those  involving  different
inalienable  noun  classes)  that have an import on the
overt grammatical properties of linguistic forms.

     I will start by discussing the general characteri-
zation of the concept of possession as it is understood
in Cognitive Grammar (CG) (Langacker  in  preparation).
In  section  3, I will discuss the formal ramifications
of the semantics of the inalienable/alienable  distinc-
tion  as  well as of other finer semantic distinctions.
Section 4 deals with the semantics of  each  possessive

2. A Schematic Characterization of Possession

     In general, possessive relations contain two enti-
ties  that  are related: one is usually called the pos-
sessor (PSR) and the  other  is  called  the  possessum
(PSM).   The  relationship that holds between these two
entities, however, is much more diverse than their  la-
bels  imply.   Namely, there are several kinds of rela-
tions that hold between them in  addition  to  "posses-
sion"  and  "ownership".  Characterizing the meaning of
what is linguistically encoded as possession  obviously
requires a more general concept than that of ownership.

     According to Seiler (p.90), "linguistic possession
consists  of  the  relationship between a substance and
another substance" in which the PSR  is  prototypically
"+animate, +human and +EGO or close to speaker" and the
PSM is "+ or -animate".  An  abstract  characterization
such  as  Seiler's is general enough to include all the
different senses of possession but it could  be  criti-
cized  as  too  extreme.   According  to  Langacker (in
preparation), a characterization of possession that al-
lows eithe participant to have either role in the rela-
tion, is so abstract that it fails to  capture  certain
asymmetries  between the PSR and the PSM.  This charac-
terization, gives no principled explanation of why  the
whole  in  part/whole relations is generally encoded as
the PSR and the part as the PSM instead of the reverse.
Similar  asymmetries can be observed when the two nomi-
nals that enter the relationship are a  person  and  an

     In the same  work,  Langacker  gives  a  schematic
characterization  of  possession  which  is  also  very
abstract but more specific than Seiler's. According  to
him,  possession is a relationship in which the speaker
"locates" one entity in relation to  another, that  is,
establishes  mental  contact.   One of the two partici-
pants in the relation is taken as a reference point  to
establish this contact with the other participant. Lan-
gacker proposes what he calls the reference-point model
(RP) to characterize possession.

     Underlying the RP model is  the  conception  of  a
world  which is populated by a variety of objects, some
of which are relatively more noticeable or salient than
others.  For example, a whole is salient in relation to
its parts; a physical object has more salience than  an
abstract  entity,  and  similarly,  a person is salient
with respect to an object.  The essential idea  of  the
RP  model  is  that non-salient objects are more easily
located in terms of salient ones than  vice  versa.   A
viewer  will  locate  a non-salient object by directing
his attention to the most readily available salient ob-
ject  first,  and from there locate the non-salient ob-
ject. The essentials of this model  are  diagrammed  in
Figure 1 below:

        <Figure 1>

     The RP model includes the following  elements:  i)
the  Reference  Point (RP), which is the salient object
in the World (W) in whose "vicinity" a non-salient  ob-
ject  is  "located";  ii)  the Target (T), which is the
non-salient object that is being located; and  the  Do-
minion  (D)  of  an RP,  which is defined as the set of
objects that can be located in terms  of  the  RP.  The
dashed  arrow  represents  the  path  through which the
viewer (V) establishes mental contact with the  target.
An  abstract characterization of possession in terms of
the RP model equates the PSR with the reference  point,
and the PSM with the T. Langacker's characterization is
abstract enough to accommodate  the  large  variety  of
senses  that  are encoded with possessive constructions
and, at the  same  time,  accounts  for  the  asymmetry
between PSR and PSM.

     Having explained the conceptual framework  assumed
in  my  analysis,  I will now turn to the data.  I will
begin  by  discussing  formal  ramifications   of   the
inalienable/alienable  distinction  as well as of other
finer semantic distinctions in Guarani in section 3.

3. Inalienable vs. Alienable Possession

     Even though Guarani does not have special morphol-
ogy  to  distinguish  alienable  and inalienable nouns,
some formal aspects indicate that there is such a  dis-
tinction.  First,  in the case of body-part and kinship
terms, the inherent character of inalienable possession
is  reflected  in the fact that these nouns do not gen-
erally occur without a specified possessor.   The  same
is  true for other nouns that designate objects or con-
cepts intimately associated with human beings, such  as
tera 'name', and ao 'clothes'.  Two examples follow:[1]

     1)  Maria o-hupi cheve i-po/*po.
         Maria 3s-lift to:me 3s-hand
        'Maria waved her hand at me'.

     2)  Maria memby/*0 memby o-mano vare'a-gui.
         Maria offspring      3s-die hunger-from
        'Maria's offspring died of hunger'.

     As can be seen in the starred  forms  of  (1)  and
(2),  the  nouns  for  "hand" and "offspring" cannot be
used without a specified PSR.  Alienable nouns, on  the
other  hand, are generally used without a possessor un-
less the speaker wants to point specifically to a  pos-
sessive relationship.

     Another    formal     manifestation     of     the
inalienable/alienable  distinction is the fact that in-
alienable possession cannot be expressed with a verb of
possession  while  alienable  possession  can. The verb
reko 'have' is one possible way of  encoding  alienable
possession  at  the clausal level.  Contrast 3b and 4b,
which represent normal expressions of alienable posses-
sion  with the ungrammatical sentences in 3a and 4a, in
which inalienable possession is encoded with a verb  of
possession.  I  will come back to this issue in section
4.4, where I discuss verbal possession.

             Inalienable                 Alienable

 3a)  *Che a-reko petei memby.    b) Che a-reko petei mesa.
       I   1s-have one offspring     I  1s-have one  table
      'I have a son'.               'I have a table'.

 4a)  ?Che a-reko petei aka.       b) Che a-reko petei ajaka.
       I  1s-have one  head          I  1s-have one  basket
      'I have a head'.              'I have a basket'.

     In the case of the inalienable examples, a kinship
term  cannot be encoded with a verb of possession under
any circumstances. Body-part terms, as in example (4a),
can be used only in very unlikely situations, for exam-
ple, when one is physically holding a head. Of  course,
in  such a case, the type of possession expressed is no
longer an inalienable one.

     Still  another   formal   manifestation   of   the
inalienable/alienable  distinction  can  be seen in the
restrictions to which "possessor ascension" is subject.
In this case only body-part terms are affected.  I will
discuss this issue at length in section 3.2.

     So far we have seen that there are some  interest-
ing  formal  indications that the alienable/inalienable
distinction has an important status in the language.  I
will now discuss basic differences in senses of posses-
sion resulting  from  the  different  inalienable  noun
classes involved.

3.1.  Possession and Different Inalienable Noun Classes

     Before I discuss the grammatical import of the se-
mantics  of the different classes of inalienable nouns,
a few words are in order about their  semantic  differ-

     In a possessive expression  containing  a  kinship
term,  both  nominals designate human beings.  The kin-
ship term is understood as the PSM and the other as the
PSR.  Kinship terms are inherently relational; that is,
they must be understood in relation  to  another  human
being.  Their relational nature makes them natural can-
didates to be encoded as PSMs since  they  have  to  be
"located" or identified in relation to another person.

     The meaning of possession involving kinship  terms
designates  an interpersonal relation. Thus, chesy 'my
mother' is not someone who is owned by me,  but  rather
someone  who  stands to me in a certain relation within
the  network  of  kinship  relations.   One  difference
between the possession involving kin relations and that
of body-parts is that in the former the PSM  is  physi-
cally  separable from the PSR while in the latter it is

     In the case of body-part nouns in a possessive re-
lation,  the  relation  between  PSM and PSR is that of
part-whole and is characterized by  physical  insepara-
bility.  According  to  the RP model, the whole is more
salient than the part, and is therefore encoded as  the
reference  point.  Possession of a body-part can be di-
agrammed as follows:

             <Figure 2>

     Here, the PSR (i.e.,  the  person)  is  identified
with  his/her body, which is encoded as an RP by virtue
of its salience with respect to its parts. The Dominion
(i.e.,  the set of objects that can be located in terms
of the RP) and the RP are also conceptually undifferen-
tiated.  The RP, in this case, is inherently made up of
distinctive  subparts  and  the   possessive   relation
results from the T being a subpart of the RP itself.

     The question now arises, do the  semantic  differ-
ences among the different subclasses have any import on
the grammatical organization of the language?  In  sec-
tion  3.2,  I  discuss  a  case  in which a grammatical
structure is clearly based on the semantics of the noun
class that it affects.

3.2.  The Semantic Basis of "Possessor Ascension".

     I will now discuss a phenomenon that  is  commonly
referred  to  as  PSR  ascension  or  PSR promotion. In
Guarani, PSR ascension co-occurs with object incorpora-
tion.  I  will start by discussing object incorporation
first.  The object of a transitive verb can be  "incor-
porated" into the verb stem (i.e. it appears as a bound
morpheme rather than a free form).  In this case, there
is  an  alteration  in the position of the nominal with
respect to the verb:  while the object normally follows
the verb, it is always preposed to the left of the ver-
bal root when it is  incorporated.  Corresponding  non-
incorporated and incorporated structures are shown in 5
(a) and (b) respectively:

    (5a) A-johei che-rova.       (b) A-j-ova-hei.
          1s-wash  I-face           1s-REF-face-wash
          'I wash my face'.         (I face-wash myself).
                                     'I wash my face'.

     In general, only unmodified nouns  can  be  incor-
porated.  Thus a noun that is modified by a determiner,
possessive or an adjective  cannot  usually  be  incor-
porated.  When the object is a body-part term in a pos-
sessive relation, however, the head noun of the  object
NP  can  be incorporated. In this case the possessor is
stranded; that is, the possessor  is  not  incorporated
along with the body-part term; instead it is encoded as
the direct object of the resulting compound stem:

(6a) Che a-johei nde-rova.      (b) Che ro-hova-hei.
     I 1s-wash  you-face            I  1s/2o-face-wash
    'I wash your face'.             (I face-wash you).
                                    'I wash your face'.

(7a) Maria oi-pete pe-mita po.  (b) Maria oi-po-pete pe mita.
     Maria 3s-slap that-kid hand    Maria 3s-hand-slap that-kid
    'M. slapped that kid's hand'.  'M. hand-slapped that kid'.

     When the PSR in the unincorporated structure is  a
pronoun,  a  pronominal  object form is prefixed to the
derived verb, as shown in (6b). When  the  PSR  of  the
unincorporated  structure  is  a lexical noun, it fills
the object slot in the incorporated structure (i.e.  it
follows the verb), as shown in (7b).

     In formal theories, examples such as (5), (6)  and
(7)  are  often  treated  as  originating from a common
underlying structure and as being related by a  deriva-
tional  process  which  involves  "possessor ascension"
(cf.  Baker, 1988 pp. 76-129). A derivational  analysis
assumes that both types of expressions are semantically
equivalent and that therefore, they must share a common
underlying  form.   Since the meaning is said to remain
constant,  the  derivation  manipulates   only   formal
categories.  According to this account, structures such
as the ones in the (a)-examples (which I will  call  A)
serve  as  the  basis  for the derivation of structures
like the ones in (b) (which I will call B). The deriva-
tion  then  would roughly consist of the following kind
of process: i) the direct  object  of  A  "becomes"  an
internal  argument of the verb and leaves the place for
the object vacant; ii) The PSR of the object  in  A  is
"promoted"  or  "ascends" to the vacant object position
of the verb in B.

     It has been pointed out that a strictly  syntactic
account    of   PSR-ascension-type   constructions   in
languages other than Guarani is inadequate (cf.  Tuggy,
and Croft, 1985). I will show that the same can be said
for the  present  data.   My  first  objection  to  the
strictly  syntactic  account  of PSR ascension concerns
the assumption that  the  meanings  of  pairs  such  as
(5),(6)  and  (7) are the same. The two types of struc-
ture in each pair  have  slightly  different  meanings.
This  claim  is corroborated by the fact that these two
types of structure are functionally different, that is,
there  are situations in which one can be used, but not
the other.  The (a)-examples have the connotation  that
the  person  is  not affected by the action to the same
extent as in the (b)-examples.  Since one generally  is
affected  by  what is done to a part of one's body, the
(b)-type structure is  most  commonly  used.  The  (a)-
structure  would  be  appropriate only in very limited,
marked circumstances.  For example, (6a) would  be  ap-
propriate to say to someone whose face one washed while
he/she was unconscious. Example (7a) is appropriate  in
situations  in  which  the  person  who slaps the kid's
hands does it accidentally; (7b), on  the  other  hand,
implies  that  the kid whose hands are slapped is being
punished, and thus affected not  only  physically,  but
socially  as  well.  In a strictly syntactic account of
possessor ascension there would be  no  principled  ac-
count of the clear semantic differences between the two
types of structures.

     Another  objection  to  the   strictly   syntactic
analysis  is  shown  in the following examples.  If all
that is involved were the promotion  of  the  syntactic
category/argument  PSR, then we would expect the PSR of
(8a) and (9a) to be coded as the clausal object in (8b,
9b), but that is not the case:

(8a) Che a-nami nde-rymba-vaka. (b)*Che ro-rymba-vaka-ami.
     I  1s-milk you-domestic-cow    I  1s/2o-domestic-cow-milk
    'I milk your cow'.             'I cow-milk you'.

(9a) Che ai-nupa ne-memby.      (b)*Che ro-memby-nupa.
     I  1s-beat you-offspring       I  1s/2o-offspring-beat
    'I beat your son'.            'I son-beat you'.

     In the examples above the PSR cannot  be  "promot-
ed"; not even in example (9b), where the possessive re-
lationship is of an inalienable nature.  Significantly,
the  only  kind  of  possessor  that  can be encoded as
clausal object in incorporated structures is  the  pos-
sessor  of  a body-part.  Semantically this makes sense
since the relationship that exists between the PSR  and
a  body-part  term is that of part to whole.  It is na-
tural then that the whole be affected whenever the part
is.  An account of Guarani possessor ascension which is
not semantically based will have no way  of  predicting
the  different behaviors of the two classes of inalien-
able nouns, which are shown here to pattern differently
with  respect  to possessor ascension.  A strictly syn-
tactic  account would need to  have  arbitrary  lexical
restrictions,  such  as  +body part, which suspiciously
happen to line up with semantic classes.

     In  this  section,  I  have  shown  that  semantic
differences in possession are reflected in the grammat-
ical organization of the language, and that a  strictly
syntactic  analysis of PSR ascension cannot account for
the restrictions to which PSR-ascension structures  are
subject  without  arbitrary  and  otherwise unmotivated
lexical  specifications.  In  CG  and  functional   ap-
proaches,  however,  meaning  differences  are not only
spelled out but are also expected, and the  grammatical
restrictions  are  given a natural explanation based on
the semantics of the structures involved. In  the  next
section,  I  will  analyze  the different ways in which
possession is encoded in Guarani and the semantic basis
of each.

4. Possession: syntactic and morphological encoding  in

     There is a variety of ways in which possession can
be encoded in Guarani.  Consider the following sample:

  (10a) Maria ajaka        (b) che-ajaka
        Maria basket            I-basket
       'Maria's basket'        'my basket'

  (11) Che che-ajaka.       (12) Che a-reko petei ajaka.
        I   I-basket             I  1s-have one  basket
       'I have a basket'.       'I have a basket'.

     I will refer to the structure in (10a) as NN  jux-
taposition,  the one in (11) as predicative-possession,
and the one in (12) as verbal possession. Given the as-
sumption of CG that grammar is symbolic, it is expected
that these different encodings of possession will  have
different  meanings. In what follows, I will character-
ize the semantic basis of each one of these structures.

4.1. N N Juxtaposition

     When the PSR is a lexical  noun  such  as  example
(10a),  nominal  possession is regularly indicated by a
simple juxtaposition of the two nominals  involved:  N1
(PSR)  N2  (PSM).  There is no special morpheme marking
the PSR or the PSM.  The  designated  (i.e.,  profiled)
element is the PSM.

     Since (10a) is a complex expression, it is reason-
able  to  investigate  how each of the elements contri-
butes to the meaning of the whole  nominal  expression.
In  Guarani, we have no overt indication that the rela-
tion between the two  nominals  is  a  possessive  one.
Taking  the  two  elements in (10a) separately, "Maria"
profiles a person and  "basket"  profiles  a  household
item.  Which  element is responsible for the possession
sense of the expression?  Since there  is  no  morpheme
responsible  for  conveying possession one might specu-
late that the possessive construction itself  is  mean-
ingful.   In  other words, the syntactic arrangement of
the two elements is responsible for the  interpretation
of  the  first  element as possessor, and the second as
possessum.  This hypothesis accords with the  claim  of
CG  that  grammatical  constructions are meaningful. We
can further speculate  that  the  NN  construction  ac-
tivates  the  RP  model in the speaker's mind and that,
within the model, he/she evokes the potential roles  of
the entities involved in the relation.  It seems, then,
that the possessive interpretation of juxtaposed  nomi-
nals  in  Guarani  is achieved by means of the RP model
plus a minimal overt clue in the construction: juxtapo-
sition  of  the two nominals, where the reference point
precedes the target (PSR-PSM).

     When the possessor is a pronoun,  possessed  nouns
regularly  display  possessive  prefixes  (see  example
(10b). I will assume that these prefixes have the  same
conceptual  content  as  the free pronominal forms, and
that the Pron-N structures are a special case of  nomi-
nal  juxtaposition.  In  addition  to  the phonological
identity, or near  identity,  of  the  free  pronominal
forms  and  the  prefixes,  the Pron-N form and the N-N
structure are clearly parallel since they have  identi-
cal linear order (i.e. the PSR precedes the PSM). Given
this parallelness, it would be unreasonable  to  assume
that the two structures are conceptually different.

     Having discussed the meaning  of  possession  when
encoded  nominally,  I will now turn to clausal posses-

4.2. Predicative Possession

     Before I present the  relevant  Guarani  data  for
this  section,  I  will  briefly  discuss the notion of
setting-subjects (see Langacker 1987),  which  will  be
relevant for my analysis of this set of data. According
to Langacker, part of the  speaker's  conceptualization
of  events  is  that the participants involved are in a
given setting.  Prototypically, one of the participants
is  selected as the clausal subject, and the setting is
expressed by an adverbial. Languages  allow  deviations
from such a prototypical manner of encoding in order to
accommodate the speaker's communicative needs. One such
departure  is  the  case of setting-subjects. In such a
case, a setting, rather than a participant, is  encoded
as  the  clausal subject.  The following examples taken
from Langacker (1987) illustrate sentences that take  a
setting for subject:

         (15) Near the fire is warm.

         (16) There are some llamas in Peru.

         (17) My cat is crawling with fleas.

         (18) The garden is swarming with bees.

         (19) noo=p no-te?       tiiwu-q.    (Luiseno)
              I=3s   my-stomach    hurt-TNS
              'I have a stomach ache'.

         (20) noo=p  no-puus  konoknis.
               I=3s  my-eye   green
              'I have green eyes'.

     In (15), the phrase headed by the locative  prepo-
sition behaves as a noun phrase and is construed as the
clausal subject. Sentence (16) contains the English ex-
istential "there", which is usually analyzed as a "dum-
my" subject. Langacker proposes to analyze  it  instead
as a "maximally schematic setting".

     In sentences (17) and (18), we have cases in which
the  settings of the events (i.e., "cat", and "garden")
are encoded as the clausal subjects  even  though  they
are  not  doing the crawling or the swarming; "cat" and
"garden" are the settings  in  which  the  events  take
place.  They  are construed as clausal subjects because
"cat" and "garden" are conceptually more prominent than
the  entities  doing the "crawling" and the "swarming".
We can say that the setting-subjects here are taken  as
reference points at the clause level in order to locate
the process of "crawling" and "swarming".

     Similar treatment is given  by  Langacker  to  the
Luiseno  sentences  in  (19) and (20); the initial pro-
nouns, which function  as  clause-level  subjects,  are
said  to specify the setting for the relation involving
the body part.  Again,  we  can  say  that  the  person
(specified  by  the  subject pronoun) is encoded as the
subject because of its  greater  conceptual  prominence
than  the  body part. Thus, in addition to being a set-
ting, it is the RP for the relation involving the  body
part. Guarani has a very similar construction involving
body-part terms:

  (21) che che-py'a-rasy.       (22) che che-resa-rovy.
       I  I-stomach-sick             I   I-eye-blue
      'I have a stomach ache'.      'I have blue eyes'.

     As in the Luiseno sentences, the pronouns  at  the
beginning  of  the  sentence  are  the clausal subjects
here, and can be analyzed as the settings for the rela-
tion  involving  the  body-parts.  In  all the examples
given above, the settings (which are  encoded  as  sub-
jects)  are  of  varying  degrees of abstractness.  The
Luiseno and Guarani examples are not "locations" in the
same physical sense that the subjects in examples (15),
(17) and (18) are; a person can only be interpreted  as
an abstract setting.

     Now, consider the following examples:

   (23) (Che) che-ajaka.
         I    I-basket
        'I have a basket'.

   (24) Che che-memby-ta.     (25) Che che-memby-se.
        I  I-offspring-FUT         I  I-offspring-DESID
       'I will have a child'.     'I want to have a child'.

     The structure of the possessive sentences above is
very similar to the ones in (21) and (22). The indepen-
dent pronominal element is the clausal  subject,  which
as in examples (21) and (22) is not the active partici-
pant subject which active verbs usually  have.  In  the
possessive  sentences,  however, the subject can hardly
be interpreted as a setting-subject for the  possessive
relation;  there  is no sense in which "I" is a setting
for the possessive relation "my offspring",  for  exam-
ple.  I  propose to analyze the subjects of the posses-
sive sentences as reference point subjects with respect
to which the possessive relation is to be understood. I
have said before that the setting-subjects can also  be
interpreted  as  a kind of RP at the clause level. Lan-
gacker  has  suggested  (personal  communication)  that
setting-subjects  are  a  special  case of RP. That is,
setting-type subjects get  generalized  into  reference
point-type  subjects. This would explain the structural
similarities between the  Guarani  sentences  given  in
(21)  and  (22)  and the possessive ones given in (23)-

     Predicative possessive sentences  do  not  have  a
verb  (which in Guarani is marked by a different set of
subject agreement markers), but have a predicative nom-
inal  as  clausal  head.  A predicative nominal results
from the temporalization of a stative  relation.  In  a
predicative  possessive structure, the possessive rela-
tion is clearly stative since it does not  designate  a
process.   Its  temporalization  is  attested to by the
fact that it can take tense markers (example  (24))  as
verbs do. Other affixes that they can take and that are
usually associated with verbs are modal affixes  (exam-
ple  (25)).  The  predicative  PSR-PSM  unit profiles a
non-verbal possession relation in the same way its nom-
inal  counterpart does, with the addition of a temporal
profile, by virtue of which the unit functions predica-

     In order to have a  possessive-predicate  meaning,
the  subject must be the same as the nominal possessor.
There is a sense in which the clausal subject is a  to-
pic.[1] This accords with a suggestion made by Langack-
er that a PSR is like a "local topic".  Since  the  PSR
here is used as a RP at the clause level, it is natural
that it is interpreted as a clausal topic.

     In the next section, I will  discuss  a  different
type  of  sentential  possession, the structures I term
verbal possession.

4.4. Verbal Possession

     This section deals with the way verbal  possession
is  encoded  in  Guarani.  The  PSR  is  encoded as the
clausal subject and the PSM  as the object.  The  mean-
ing  of  such sentences is equivalent to the meaning of
English sentences of the type:"NP have NP":

    (27) Che a-reko petei mita.[1]
         I   1s-have one child
         'I have a child'.

     The PSR and the PSM are encoded  as  participants.
The  PSR is construed not as a mere RP, but as a parti-
cipant exerting energy on the PSM, which is encoded  as
the object. Energy is exerted in the sense that the PSR
"holds" the PSM in its dominion. It is important to no-
tice  in  this regard that the verb reko also means 'to
physically hold something', as in:

         (28) A-reko petei kyse che-po-pe.
              1s-hold one  knife I-hand-in
              'I have a knife in my hand'.

     The verb  reko  does not convey  intimate  posses-
sion; it usually designates transient, impermanent con-
trol of the PSR on the PSM. Thus (29) below is perfect-
ly appropriate:

         (29) A-reko Maria mesa che-roga-pe.
              1s-have Maria table I-house-in
              'I have Maria's table at home'.

     In the example above,  reko  refers  to  temporary
control;  it does not convey permanent possession since
the table belongs to Maria. Given  the  fact  that  the
verb  reko  does  not convey intimate possession, it is
natural that possession of kinship and body-part  terms
is not usually encoded with this type of structure (see
examples in section 3 on the inalienable/alienable dis-

     Predicative possessive sentences,  as  illustrated
in  (23)-(25), do not mean 'NP have NP' in the same way
that verbal possessives do. Predicative possessives ex-
press  a more intimate and permanent type of possession
than does verbal possession.  Thus  regardless  of  the
class  of  nouns  involved (alienable/inalienable), the
two different linguistic structures express  a  differ-
ence  in the degree of intimacy or separability between
the PSR and PSM. Consider the following examples:

      (30)  Che che-roga Paraguay-pe.
            I   I-house Paraguay-in
           'I have a house in Paraguay'.

      (31) Che a-reko petei oga Paraguay-pe.
            I  1s-have one  house Paraguay-pe
           'I have a house in Paraguay'.

     Example (30) above is used in cases in  which  the
PSR  owns  a house and lives there permanently. Example
(31), however, has the connotation that the person owns
the  house  but  does not live there permanently.  This
difference in meaning between the two types  of  struc-
ture is reflected formally in the fact that the PSR and
the PSM are not separated by a verb in predicative pos-
sessives,  while  in verbal possessives the PSR and PSM
are separated by a verb.

     Having described the meaning of the different ways
in which Guarani encodes possession, I will end by dis-
cussing the functional complementarity of the different
structures.   I  will  use  notions  such  as  Seiler's
linguistic continuum and Haiman's linguistic iconicity.

5. The Function and Iconicity of Possessive Structures

     According to Seiler (1983) and Brettschneider  and
Seiler (1985), the motivation  underlying the structur-
al diversity with which possession is  encoded  is  the
existence  of  two properties that play a complementary
role: inalienable  vs.  alienable.   Possessive  struc-
tures,  they say, are ordered along a continuum accord-
ing to the degree of structural complexity.  They claim
that  the  more complicated the structure, the more ex-
plicit is the type of relation between the PSR and  the
PSM (the extreme case of explicitness is a verb of pos-
session). "Intimate" possession does  not  need  to  be
linguistically explicit; therefore, inalienable posses-
sion tends to be encoded with less explicit  linguistic
means.  The extreme case of inexplicitness is NN juxta-
position. On the other hand, the less consistently  the
object  belongs to the sphere of intimacy, the stronger
is the need to make  the  relationship  explicit.  Thus
there  is a tension between ways of encoding possession
due to the presence of two opposite  concepts,  aliena-
bility  vs.  inalienability, which play a complementary
role in the language in such a way that the increase in
participation in one category implies a decrease in the

     Seiler's findings fit very well with a  more  gen-
eral  picture of linguistic iconicity developed by Hai-
man (1983).  According to Haiman, linguistic structures
are  iconic  in the sense that they tend to reflect the
type of conceptualization they stand for. Haiman's gen-
eralizations  are  as follows: "the linguistic distance
between expressions corresponds to the conceptual  dis-
tance  between  the  ideas  they  represent",  and "the
linguistic separatedness of an  expression  corresponds
to  the  conceptual independence of the object or event
which it represents" (pp. 782-783).

     Let us return now to some  facts  about  the  dif-
ferent  possessive structures in Guarani and re-examine
them in the light of these generalizations.

     First, the claim that verbal  possession tends  to
designate alienable possession holds for Guarani. As we
have seen, possession of body-part and  kinship  cannot
be encoded with a verb of possession. Taking the iconi-
city generalization, the explanation would be  as  fol-
lows:  in  verbal  possession,  the PSR and the PSM are
separated by a verb.  This structure is awkward because
it  does not reflect the intimate or inseparable nature
of the relation between the PSR  and  the  PSM.   Also,
body  parts  and kinship terms are not conceptually in-
dependent, since they are generally conceived in  rela-
tion  to something else. Encoding them as separate from
their PSRs does not reflect their conceptual dependency
and the result is strange.

     As far as NN  juxtaposition  and  the  predicative
possessive  are concerned, there is no restriction with
respect to the class of nouns that can be encoded  with
these  structures.  Thus both alienable and inalienable
nouns can be encoded as PSMs in  juxtaposition  and  in
predicative  possessives.   If  these  structures do in
fact represent an intimate  relation  between  PSR  and
PSM,  this suggests that alienable nouns, despite their
separable nature, are easily construed as  being  inti-
mately associated with the PSR. Inalienable nouns, how-
ever, are more resistant to being encoded as separable.

     Despite the fact that alienable nouns are  concep-
tually separable, they can be construed as more or less
intimately associated with the PSR in a possessive  re-
lation  via the choice of linguistic structure. Thus we
saw in section 4.4 that there is a  difference  in  the
degree of intimacy of the possessive relation conveyed,
depending on whether or not  the  relation  is  encoded
with  a verbal possessive structure (see examples (30)-
(31). Thus, a nominal predicative structure  conveys  a
more permanent and close type of relationship by virtue
of the fact that it  has  a  less  explicit  linguistic
structure.   A  verbal  possessive structure, which in-
volves a more explicit linguistic encoding,  conveys  a
less intimate type of possession.

6. Concluding Remarks

     In this paper I have shown that:  i)  despite  the
absence of a specific morphology signaling the distinc-
tion inalienable/alienable in Guarani, this distinction
does  have an important status in the language as indi-
cated by its formal ramifications.   I  have  suggested
that  these  two  opposite  concepts  underlie the com-
plementarity of the different possessive structures  in
the language.

     ii) Apart from the inalienable/alienable  distinc-
tion  there  are  finer  semantic  differences (such as
those  involving  different  semantically  based   noun
classes)  that  have an effect on the overt grammatical
properties of linguistic forms. In particular,  I  have
shown  that  the formal restrictions pertaining to pos-
session ascension structures are semantically motivated
and  that a strictly syntactic analysis does not have a
natural way of explaining their formal restrictions be-
cause  it wrongly assumes semantic identity between PSR
ascension paraphrases.

     iii) It was shown that while a nominal predicative
structure expresses an intimate, permanent type of pos-
session, verbal possession conveys a less  intimate  or
permanent   possessive  relation.  This  suggests  that
linguistic structure alone can convey a  difference  in
the  degree  of  intimacy or permanency of a possessive
relation. Thus the claim of CG that  linguistic  struc-
tures are symbolic and meaningful is strongly supported
by the present data.


I wish to  express  my  gratitude  to  Suzanne  Kemmer,
Ronald  Langacker, Rick Floyd and Ricardo Maldonado for
their constructive criticism of earlier drafts of  this


1- The following are some  of  the  abbreviations  that
will be used in the glosses:

              SUP  =  superlative
              REF  =  reflexive
              TOT  =  totalitative ("completely", "all")
              PASS =  passive
              DESID=  desiderative

2- I am using the term "topic" here in  the  functional
sense. That is, the element which is the "center of at-
tention" and which "specifies the domain  within  which
the  predication  holds" or "announces the theme of the
discourse" (see Li and Thompson, 1975 p. 464). Used  in
this sense, of course, there is a great deal of overlap
between the notions subject and topic.

3- The word mita 'child', unlike memby in example  (24)
above, is not a kinship term. In other words, it is not
in opposition to the words for  'mother'  or  'father',
but to the word for 'adult'.


Allen, Sidney.  1964.  "Transitivity  and  Possession".
     Lg.  40:337-43.
Baker, Mark C. 1988.  Incorporation. A Theory of Gram-
     matical  Function Changing. The University of Chi-
     cago Press.
Brettschneider, Gunter and  Seiler,  H.  (eds.).  1985.
     Language  Invariants  and  Mental Operations. Tu-
     bingen: Gunter Narr Verlag.
Croft, William. 1985. "Indirect Object 'Lowering'". BLS
Givon, Talmy. 1975. "Topic,  Pronoun,  and  Grammatical
     Agreement".  In  Subject and Topic, ed. by Charles
     Li. Academic Press.
______ 1984. Syntax: A Functional Typological Introduc-
     tion.  John Benjamins Publishing Company.
Haiman, John. 1983. "Iconic and  Economic  Motivation".
     Lg.  59:781-819
Hyman, Larry et al. 1970.  "Inalienable  Possession  in
     Igbo". J.W.A.L.. 7,2:85-101.
Langacker, Ronald. 1987. Foundations of Cognitive Gram-
     mar.   Vol.  1,  Theoretical Prerequisites.  Stan-
     ford: Stanford University Press.
______ 1986.  "Settings, Participants, and  Grammatical
     Relations". Proceedings of the Second Annual Meeting 
     of the Pacific Linguistcsi Conference.
______  1987.   "Grammatical   Ramifications   of   the
     Setting/Participant Distinction". BLS 13:383-94.
______ In prep. Foundations of Cognitive Grammar.  Vol.
Li, Charles and Thompson, Sandra.  1975.   Subject  and
     Topic: A New Typology of Languauge. In Subject and
     Topic, ed. by Charles Li.  Academic Press.
Sapir,  E.  1917.   "Review  of   Het   Identificeerend
     Karakter  der  Possessieve  flexie  in  Talen  van
     Noord-America by C.C. Uhlenbeck".  IJAL 1:86-90.
Seiler, Hansjakob. 1983. "Possessivity, Subject and Ob-
     ject".  Studies in Language. 7(1): 89-117.
Steele, Susan. 1977. "On Being Possessed".  BLS  3.114-
Taylor, John R. (unpublished paper). "Possessive  Geni-
     tives in English".
Tuggy, David. ms. "Ethical Dative and  Possessor  Omis-
     sion si, Possessor Ascension No!".

[A similar verion of this paper was presented at the 
Fifth Annual Meeting of the Pacific Linguistics Conference, 
Eugene Oregon, May 1989.  Hard copies of the paper, including 
figures may be requested from CRL.]