INTRODUCTION

The Apurinã language is spoken by the Apurinã people living along the tributaries of the Purus River in the western Amazon region of Brazil. What I will present here is a particular type of categorization of nouns.

The problem consists of the following: as seen in the example la., a possession construction can be simply marked by the juxtaposition of nominal elements. In the example la. the noun kywy 'HEAD' is interpreted as referring to a possessed element, while the preceding nominal kyky 'MAN' refers to the possessor element. In the following example in 1b., kyky again refers to the possessor, and amaro-te 'DAUGHTER' refers to the possessed element. The relevant distinction between the two possession constructions in la. and 1b. is that in the second one the head noun referring to the possessed element is marked by the suffix -te. In 1c. and 1d. the head nouns kota-re 'BASKET' and ããta-ne 'CANOE' are also suffixed, but by different morphological elements. My question is whether there is any plausible motivation for the Apurinã speakers to keep morphological markers in the language to mark nominal possession when a possession construction can be identified by juxtaposition.

la. [kyky kywy] n-etama

MAN HEAD 1SG.SUBJ-SEE

'I SAW THE MAN'S HEAD.'

b. [kyky amaro-te] n-etama

MAN DAUGHTER-POSSED 1SG.SUBJ-SEE

'I SAW THE MAN'S DAUGHTER.'

c. Mipa apokar-ry [syto kota-re]

MIPA FIND-3SG.OBJ WOMAN BASKET-POSSED

'MIPA FOUND THE WOMAN'S BASKET.'

d. owa apoka-ry [Mipa ããta-ne]

SHE FIND-3SG.M.OBJ MIPA CANOE-POSSED

'SHE FOUND MIPA'S CANOE.'

POSSESSION AND LINEAR WORD ORDER

A possible answer to the question of why have morphological possession markers in the first place can be postulated as a matter of ambiguity. The examples in 2. show that, for some constructions in the language, linear word order cannot distinguish ATTRIBUTIVE from POSSESSION modification. In this case, the morphological marker distinguishes two syntactic constructions. While in 2a. the head element of the nominal construction kema syto 'FEMALE TAPIR,' is interpreted as an attributive modifier, in 2b. the same head element, when marked by the suffix -re, is interpreted as the possession modifier. Similar examples are given in 3a. and 3b.

2a. ynawa nyta-ro [kema syto]

THEY LOOK.FOR-3F.OBJ TAPIR WOMAN

'THEY LOOKED FOR A FEMALE TAPIR.'

b. ynawa nyta-ro [kema syto-re]

THEY LOOK.FOR-3F.OBJ TAPIR WOMAN-POSSED

'THEY LOOK FOR THE TAPIR'S FEMALE.'

3a. ynawa nyta-ry [kema kyky]

THEY LOOK.FOR-3M.OBJ TAPIR MAN

'THEY LOOKED FOR A MALE TAPIR.'

b. ynawa nyta-ry [kema kyky-re]

THEY LOOK.FOR-3M.0BJ TAPIR MAN-POSSED

'THEY LOOKED FOR THE TAPIR' S MALE.'

Therefore, there is a syntactic motivation for having morphological possession markers in the language. In order to investigate the lexical properties of the nouns that take those markers, I will distinguish the POSSESSION HEAD NOUNS from the UNPOSSESSION HEAD NOUNS -- hereafter I will call the latter ABSOLUTES (influenced by Seiler 1983). Absolutes are nouns that cannot occur as the possessed head element in a possession construction.

As a first example, the noun ããta-ne CANOE given in 1d. is the possessed element suffixed by -ne. The noun ããta cannot occur as a possessed head noun unless it carries the appropriate morphological marker (i.e. -ne), as seen in 4. Moreover, once suffixed by -ne, the same noun can only occur as the possessed head, cf. ããta-ne in 5.

4. *owa apoka-ry [Mipa ããta]

SHE FIND-3SG.M.OBJ MIPA CANOE

(SHE FOUND MIPA'S CANOE.)

5. *owa apoka-ry [ããta-ne]

SHE FIND-3SG.M.OBJ CANOE-POSSED

(SHE FOUND THE CANOE.)

On the other hand, if ããta is not suffixed by -ne, it can occur as an absolute noun with no morphological marker, as in 6. I will refer to the nouns following the behavior of ããta as OVERT POSSESSION head nouns -- since they are morphologically marked in possession head position, or COVERT ABSOLUTES -- since they are not marked when occurring in a role other than possession head.

6. owa apoka-ry [ããta]

SHE FIND-3SG.M.OBJ CANOE-POSSED

SHE FOUND THE CANOE.

In contrast, the noun serepi 'ARROW' stands by itself in 7. as the possessed head element of the nominal construction constituting the subject of the clause,

7. [popy~kary serepi] irika

INDIAN ARROW FALL

'THE INDIAN'S ARROW FELL DOWN.'

whereas in 8. the same noun cannot stand by itself as the nominal expression constituting the same subject role in an absolute form:

8. *[serepi] irika

ARROW FALL

(THE ARROW FELL DOWN.)

Instead, as 9. shows, the addition of the suffix -txi is required for the noun to stand as the subject nominal expression in its absolute form:

9. [serepi-txi] irika

ARROW-ABS FALL

THE ARROW FELL DOWN.

Furthermore, as seen in 10., the suffix -txi also disallows the possibility of the noun to occur as the possessed head element:

10. *[popy~kary serepi-txi] irika

INDIAN ARROW-ABS FALL

(THE INDIAN'S ARROW FELL DOWN.)

I will call the nouns following the behavior of serepi COVERT POSSESSION head nouns -- since they require no morphological marker as possession heads, or OVERT ABSOLUTE nouns -- since they do require a morphological marker to occur as anything that is not a possession head noun.

There also are nouns that are different from serepi and ããta insofar as they require (distinct) morphological markers for both absolute and possessed head nouns. An example is the noun kota-re, as already illustrated in 1c. The next examples in 11. and 12. aim to show that the suffix -re is required for the noun kota to occur as a possessed head:

11. *Mipa apokar-ry [syto kota]

MIPA FIND-3SG.OBJ WOMAN BASKET-POSSED

(MIPA FOUND THE WOMAN'S BASKET.)

12. Mipa apoka-ry [syto kota-re]

MIPA FIND-3SG.OBJ WOMAN BASKET-POSSED

MIPA FOUND THE WOMAN'S BASKET.

And the following example in 13., in contrast with the one in 1c., indicates that the same construction is ungrammatical if kota stands by itself in the absolute role without the appropriate morphological marker:

13. *Mipa apokar-ry kota]

MIPA FIND-3SG.OBJ BASKET

(MIPA FOUND THE WOMAN'S BASKET.)

Following the terminology already introduced above, I will also refer to the nouns with the behavior of kota as overt possession head nouns -- as they also require morphological marker in possessed head position, or OVERT ABSOLUTE head nouns -- as they require morphological markers in non- possession head position.

As a result of the distinctions established between OVERT versus COVERT possession heads or absolutes, three patterns of nominal constructions arise out of the combination of possession plus absolute marking, listed in Table 1 as Patterns I, II, and III:

Table 1: Patterns of Possession plus Absolute Markings

	NOMINAL PATTERNS	POSSESSION	ABSOLUTE
	Pattern I	Overt (-te/-re/-ne)	Covert (-Ø)
	Pattern II	Covert (-Ø)	Overt (-txi)
	Pattern III	Overt (-re)	Overt (-ry)

In the next section, I will explore a little further the properties of each of the marking patterns.

Pattern I: OVERT POSSESSION AND COVERT ABSOLUTE (or UNPOSSESSION)

In the first marking pattern of nominal constructions, a noun takes one of a paradigm of three suffixes (-te, -re, or -ne) when this noun occurs as the possessed element, head of a nominal construction; and the same noun does not take any morphological marker when occurring in absolute constructions. In 14., for instance, the nominal construction mipa aiko-te 'MIPA'S HOUSE,' which is the subject of the clause, consists of a possessed head noun morphologically marked and a preceding nominal possessor. In 15., the noun aiko 'HOUSE' occurs with no morphological marker in its absolute form. In 16. the example shows that once the suffix -te is attached to a noun, such a noun cannot function as absolute. The examples in 17-19 follow the same pattern:

14. [mipa aiko-te] iri-pe

MIPA HOUSE-POSSED FALL-PFTV

'MIPA'S HOUSE HAS FALLEN DOWN.

15. [aiko] iri-pe

HOUSE FALL-PFTV

'THE HOUSE HAS FALLEN DOWN.'

16. *[aiko-te] iri-pe

HOUSE FALL-PFTV

(THE HOUSE HAS FALLEN DOWN.)

17. [pite ximaky-te] ata nhika

YOUR FISH-POSSED WE EAT

'WE ATE YOUR FISH.'

18. [ximaky] ata nhika

FISH WE EAT

'WE ATE FISH.'

19. *[ximaky-te] ata nhika

FISH-POSSED WE EAT

'WE ATE FISH.'

The structural properties presented for the suffix -te also apply to the other possession head suffixes -re and -ne.

The semantics of Overt Possession Head Nouns and Covert Absolutes

Considering that the three suffixes (-te -re and -ne) share identical structural properties, what could we say, then, about their semantics? Tables 2-4 list the absolute and possessed forms of sets of nouns for each of the three suffixes. A comparison of the three sets of nouns does not immediately seem to correlate with any recognizable semantic grouping. For instance, in Table 2 we can find nouns referring to long and rigid objects ('TREE, WOOD'), flat and flexible objects ('LEAF'), round objects ('BALL'), humans ('DAUGHTER, SON'), fish, etc. In Table 3, nouns may refer to plant parts ('KNOT'), humans ('MAN, WOMAN'), utensils ('SMALL PAN'), and abstract nouns ('WORK, TIREDNESS'). In Table 4, again we see nouns referring to utensils ('KNIFE, GRATER, PAN'), nature elements ('SUN, STONE'), plant names ('GRASS'), and others whose physical shape, consistency or other explicit semantic characteristics do not motivate any semantic grouping. Nevertheless, as a whole, the three sets seem to exclude KINSHIP TERMS and BODY PARTS, except for the nouns referring to 'DAUGHTER' and 'SON,' and the noun referring to 'BEAK.'

Table 2: Possession Head Marker -te

ABSOLUTE FORM	POSSESSED FORM	GLOSS
ããmyna	ããmyna-te	TREE, WOOD
ããtsopa	ããtsopa-te	LEAF
aiko	aiko-te	HUT
amaro	amaro-te	DAUGHTER
amary	amary-te	SON
i~porãã	i~porãã	WATER
itokory	itokory-te	FIELD FARM 1
jãri~jã	jãri~jã-te	BEVERAGE
jomaky	jomaky-te	BEAK
kaisoro	kaisoro-te	STONE TO SHARPEN
kanawa	kanawa-te	CANOE
kasyry	kasyry-te	MOON
kikio	kikio-te	FIELD FARM 2
kypatxi	kypatxi-te	SAND
manee	man eete	RIVER'S MOUTH
syti~i~	syti~i~-te	BALL
tipykijã	tiprkijã-te	DOOR
xamynaky	xamynaky-te	SHOTGUN
xamypoky	xamypoky-te	STOVE
ximaky	ximaky-te	FISH
etc...

Table 3: Head Possession Marker -re

ABSOLUTE FORM	POSSESSED FORM	GLOSS
jaxirika	jaxirika-re	KNOT; TO TIE
kama	kama-re	WORK; TO WORK
kopitita	kopitita-re	SMALL PAN
kyky	kyky-re	MAN
sãpaka	sãpaka-re	WEARINESS; TO GET TIRED
syto	syto-re	WOMAN
etc...

Table 4: Possession Head Marker -ne

ABSOLUTE FORM	POSSESSED FORM	GLOSS
jowata	jowata-ne	KNIFE
ããtapanhi	ããtapanhi-ne	ASH
i~topa	i~topa-ne	(AREA IN THE) JUNGLE
xiti	xiti-ne	LAND, LOCALTTY
katsotaty	katsotaty-ne	GRASS
kopiti	kopiti-ne	PAN
sawatapa	sawatapa-ne	SHOES
atokatxi	atokatxi-ne	SUN
ããta	ããta-ne	TRADITIONAL CANOE
kai	kai-ne	STONE
kipeta	kipeta-ne	INSTRUCTION TO MAKE A STRING
paraka	paraka-ne	HOUSE
tõõpa	tõõpa-ne	PLACE
etc...

Therefore, we can say that the semantics of overt possessed nouns and covert absolutes consist of a mixture of human and inanimate references, whose shape and consistency vary in unpredictable ways. However, the fact that, overall, overt possessed or covert absolute nouns partially exclude nouns referring to body parts and kinship relations suggests that the nouns involved in Pattern I refer to references that are NOT prototypical inalienable.

Pattern II: COVERT POSSESSION AND OVERT ABSOLUTE

The second pattern of possession and absolute marking consists of the nouns that are not morphologically marked as the possession head; instead, these nouns are morphologically marked when they do not occur as the possession head; that is, they are 'unmarked' as possession heads and 'marked' as absolutes. In this case, the absolute marker is the suffix -txi. Compare 20a and 20c.

20a. [syto mãka] iri-pe

WOMAN CLOTHES FALL-PFTV

'THE WOMAN'S CLOTHES HAVE FALLEN DOWN.'

b. *[mãka] iri-pe

CLOTHES FALL-PFTV

(THE CLOTHES HAVE FALLEN DOWN.)

c. [mãka-txi] iri-pe

CLOTHES-ABS FALL-PFTV

'THE CLOTHES HAVE FALLEN DOWN.'

d. *[syto mãka-txi] iri-pe

WOMAN CLOTHES-ABS FALL-PFTV

(THE WOMAN'S CLOTHES HAVE FALLEN DOWN.)

The semantics of Covert Possession Head Nouns and Overt Absolute

For this pattern, we can see transparent semantic subgroupings. In Table 5 the nouns refer to BODY PARTS, in Table 6 they refer to PERSONAL BELONGINGS, in Table 7 they refer to elements that may be seen as characteristically ABSTRACT, in 8 they are KINSHIP TERMS and, finally, in 9 they refer to PLANT PARTS. Thus, the semantics of the nouns that are covertly marked for possession and overtly marked for absolute corresponds to that of prototypical inalienable nouns. (The nouns with a question mark in Tables 5-7 are those for which an absolute form is not yet attested.)

Table 5: Absolute with - and Possessed with No Marker: BodyParts

ABSOLUTE FORM	HEAD POSSESSED FORM	GLOSS
api-txi	apy	BONE
hereka-txi	hereka	BLOOD
i~i~-txi	i~I~	FAT
kywy	kywy~-txi	HEAD
kanõ-txi	kano	ARM
kiri~-txi	kiri	NOSE
mane-txl	mane	BODY
mata-?	mata	SKIN
mekita-?	mekita	WING
noky-txi	noky	NAPE
pipi-txi	pipi	PENIS
piti-?	piti	FEATHER
pitxi~-txi	pitxi	PENIS
porike-?	porike	THIGH
poriky-?	poriky	BACK
poto-?	poto	LIP
pyty~-txi	pyty	INNER PART OF VAGINA
sokõ-txi	soko	VAGINA
tikako-?	tikako	BELLY
tõ-txi	tõ	FACE
xenhi~-txi	xenhi	MEAT
xiaky~-txi	xiaky	GENITAL HAIR
ynama-?	ynama	MOUTH
etc...

Table 6: Absolute with -txi and Possessed with No Marker: Personal Belongings

ABSOLUTE FORM	POSSESSED FORM	GLOSS
awapoko-txi	awapoko	VILLAGE
awinhi-?	awinhi	HOUSE
isawatapa-?	isawata-pa	SHOES
jokatõ-?	jokatõ	(BODY) PAINTING
keko-txi	keko	HAMMOCK
kitimata-?	kiti-mata	SANDAL
koi-txi	koi	FLUTE
mãka-txi	mãka	CLOTHES
meko-txi	meko	PADDLE
nokytsa-txi	nokytsa	NECKLACE
poky~-txi	poky	STOVE
tapo-txa	tapo	BOW
toy-txi	toy	THING
xerepi-txi	xerepi	ARROW
etc...

Table 7: Absolute with -txi and Possessed with No Marker: Abstract Nouns

ABSOLUTE FORM	POSSESSED FORM	GLOSS
hi~wãka-txi	hiwãka	NAME
jenany-?	jenany	HAPPINESS
mixi~-txi	mixi	PREGNANCY
pari~ka-txi	pari~ka	WORK,ACTIVITY
pirana-?	pirana	TALK, SPEECH
posonata-txi	posonata	THIRST
sãkire-txi	sãkire	LANGUAGE
sereka-txi	sereka	DANCE
etc...

In Tables 8-9, the nouns appear to be consistent in that they never occur as absolute. In Table 8 the nouns are kinship terms, and only their vocative (or addressing forms) are attested in absolute forms. In Table 9, the nouns refer to plant parts, and they are attested only in constructions in which the part is possessed by the whole, but NEVER when the part can stand by itself as an absolute.

Table 8: Absolute and Possessed with No Marker: Kinship Terms

ABSOLUTE	POSSESSED FORM	GLOSS
	akyro	GRANDMA
	atokyry	GRANDPA
	epyri	BROTHER
	itari	BROTHER
	omekanhiry	GRANDSON
	y~tanyro	WIFE
	y~tanyry	HUSBAND
	ykõkyry	UNCLE
	ynyro	MOTHER
	ynyro-tanyro	AUNT
	yry	FATHER
	etc...

Table 9: Absolute and Possessed with No Marker: Plant Parts:

ABSOLUTE FORM	POSSESSED FORM	GLOSS
	hi~wi~	FLOWER
	ykaty	TREETOP
	kijana	VEGETABLES
	ky	KERNEL, SEED
	pori	BRANCH
	pytsa	LIANA
	tãta	BARK
	tsomita	ROOT
	tsota	TRUNK
	etc...

Thus, Patterns I and II suggest that, to a certain degree, these morphological markers serve to distinguish ALIENABILITY versus INALIENABILITY.

Pattern III: OVERT POSSESSION AND OVERT ABSOLUTE

In the third type of nominal construction, the noun is marked by the suffix -re in possessions and -ry in absolutes, as shown in 21. and 22. These are the Overt Possession and Overt Absolute constructions.

21a. [kawy-ry] amaro nhika

POPUNHA-ABS GIRL EAT

'THE GLRL EATS POPUNHA (A FRUIT)'

b. ny-[kawy-re] amaro nhika

1SG-POPUNHA-POSSED GIRL EAT

'THE GIRL EATS MY POPUNHA'

c. *ny-[kawy-ry] amaro nhika

1SG-POPUNHA-ABS GIRL EAT

(THE GIRL EATS MY POPUNHA.)

22a. [kota-ry] syto kama

BASKET-ABS WOMAN MAKE

'THE WOMAN MADE BASKET.'

b. ny-[kota-re] syto kama

1SG-BASKET-POSSED WOMAN MAKE

'THE WOMAN MADE MY BASKET.'

c. *ny-[kota-ry] syto kama

1SG-BASKET-ABS WOMAN MAKE

(THE WOMAN MADE MY BASKET.)

The Semantics of Overt Possession Head and Overt Absolute

For the nouns which are overtly marked for possession and also for absolute, there is no clear-cut semantic grouping. As shown in Table 10, nouns in Pattern III can refer to MANUFACTURED (PERSONAL) OBJECTS ('FLUTE, STOOL, BASKET'), LIQUIDS ('JUICE'), FRUITS ('POPUNHA'), etc. of various shapes and consistency. Nevertheless, except for the noun referring to 'HIPS,' which has not yet been attested in the absolute form, there is no instance of -re and -ry occurring with body parts or kinship terms.

Table 10: Head Possession Marker -re

ABSOLUTE FORM	POSSESSED FORM	GLOSS
hãpoky-ry	hãpoky-re	FLUTE
ikijana-ry	ikijana-re	POISON
jã-ry	i-jã-re	JUICE
jotipana-rY,	jotipana-re	STOOL
kawy-ry	kawy-re	"POPUNHA" (FRUIT)
kota-ry	kota-re	BASKET
nhipoko-ry	nhipoko-re	FooD
pirike-ry	pirike-re	WALL
taka-ry)	taka-re	PLANTATION
tsowaky-?	tsowaky-re	HIPS
txipoko-ry	txipoko-re	FRUIT
xika-ry	xika-re	SONG
xoka-ry	xoka-re	FLATULENCE
etc...

Both Patterns I and III, then, are distinct from Pattern II -- which has clear semantic correlation. This distinction seems to be that between alienability versus inalienability.

POSSESSION HEAD/ABSOLUTE MARKTNG AND NOUN CLASSES

There are, therefore, three different construction types based on possession and absolute marking patterns corresponding to three attested types of NOUN CLASSES found in the language. These noun classes can be identified by the patterns of their formal marking in possession and absolute constructions, as shown in Table 11. Notice that, although four classes would be possible logically, there is NO fourth class in which the same nouns would be unmarked for both the possession head and the absolute.

Table 11: Nominal Classes

NOUN CLASS	POSSESSION	ABSOLUTE
Class I	Overt (-te/-re/-ne)	Covert (-Ø)
Class II	Covert (-Ø)	Overt (-txi)
Class III	Overt (-re)	Overt (-ry)
*(Class IV	Covert (-Ø)	Covert (-Ø) )

The possession head markers -te, -re, -ne, or the absence of a formal marking, (-Ø) resemble GENITIVE CLASSIFIERS by occurring with different nouns only in possession constructions. A synchronic approach to the phenomenon, however, suggests that, presently, these markers can be structurally distinguished from typical genitive classifiers as, for example, those of the sample of Oceanic and South American languages given in Carlson and Payne 1987, insofar as the genitive classifiers normally occur attached to the noun referring to the possessor, while the possession head markers are attached to the noun referring to the possessed element. Moreover, as in Polynesian languages (Cf. Nichols 1988:574), the possession marking of Apurinã also resembles GENDER, and, as such, can be better analyzed as a system of noun class markers rather than one of noun classifiers. In Apurinã, however, the possession marking forms a larger system of marking patterns which includes the absolute marking in categorizing nouns.

Therefore, the three distinct patterns of possession marking in the language express three noun subcategories which are sensitive to the grammatical (and perhaps cognitive) role of nominal possession in the language. Now, finally, to try to understand how the Apurinã speaker learns which nouns follow which of the three marking patterns, one can postulate that there may also be a pattern of semantic distribution of nouns in different classes in terms of three populations: the first consists of prototypical non-inalienable or most alienable like nouns; the second consists of the prototypical inalienable nouns; and the third consists of nouns which are intermediary in terms of the two former classes. This idea can be represented as in the figure below. The intersection of Patterns I and II would give rise to the Pattern III; this latter Pattern III would consist of the individual references that would not fit into the set of typical inalienable or alienable references:

Pattern Pattern Pattern

I III II

Although this hypothesis would need to be verified in a larger corpus of linguistic data, some pieces of evidence for this suggestion arise out of a consideration of the function of possession marking, its iconicity and markedness properties.

ON THE FUNCTION OF POSSESSION MARKING

As shown above, the categorization of the nouns as expressed in the patterns involving possession and absolute constructions is not based on animacy, shape or consistency of the noun references. The exclusion of such semantic features here can be motivated by the relevant functional roles of possession marking. As it happens, the association between a possessor and a possessed element tends to be based on the FUNCTIONAL or SOCIAL role of the latter in relation to the former; that is, having a paradigm of nominal possession markers available in the language, we would find that each noun category may be marked by a specific possession marker depending on the functional or social role a possessed reference has to the possessor. For example, the possession marker used for kinship terms would be different from that used for words like 'DOOR, BALL...' not because of their difference in terms of their animacy, shape or consistency, but in terms of what functionally or socially the possessed element represents to the possessor. These functional factors involved in possession marking have been shown in Carlson and Payne (1987) for possession constructions marked by genitive classifiers. To translate this in terms of in/alienability, the distinction between alienable and inalienable constructions is primarily based on functional or social relationships between the noun references involved. Therefore, at least onginally, this functional motivation may perhaps be important to understand why a possession marking system as the one found in Apurinã can be traced through the protolanguage (i.e. Proto-Maipuran), thousands of years ago (cf. reconstruction of David Payne 1993).

The function of possession and absolute marking patterns in this language is, thus, to encode alienability and inalienability. Alienability and Inalienability in Apurinã may be partially defined on socio-cultural grounds; that is, by the specific cultural and social variables that condition the Apurinã speakers to treat a set of nominal references as more alienable and another set of nominal references as more inalienable. The extent to which the Apurinã speakers are capable of identifying the cultural or social variables into play in the noun categorization system will determine whether this system is semantically or grammatically based. A deeper knowledge of the Apurinã culture and society is required by a linguist in order to access such extra-linguistic factors renected in the language. The data suggest that noun categorization in Apurinã is grammatically expressed by the patterns of possession and absolute marking; but the noun categories express a continuum which goes from prototypical alienability to prototypical inalienability. The fact that, in deciding which nouns follow each pattern, there is a recognizable semantic contrast only between the two extremes of the alienable-inalienable continuum suggests a process of semantic bleaching in the system of noun categorization; the semantic bleaching may be leading to a more grammatical, and less semantic, distinction between alienability versus inalienability.

POSSESSION MARKING AND ICONICITY

It has been argued (Seiler 1983, Nichols 1988, Chappell & McGregor 1989) that there tends to be a closer relationship between possessor and possessed references in inalienable constructions. It appears to be the case that this closer relationship found between possessor and possessed references in inalienable constructions is iconically renected in Apurinã, insofar as the prototypical inalienable expressed by Pattern II is morphologically unmarked for possession but marked for absolute. That is, if covert possession head nouns and overt absolutes express inalienable references, there is iconicity in the fact that such nouns are covert rather than overt possession head, as well as overt rather than covert absolute. In contrast, the less close relationship between possessor and possessee in alienable constructions is also iconically marked in that the constructions following Pattern I are morphologically marked in possession but not in absolute constructions. That is, if overt possession head nouns express alienable references, there is iconicity in the fact that such nouns are overt rather than covert possession heads. If the constructions following Pattern III indeed are intermediary, they would express types of associations between noun references which are not well defined as conceptually close or distant. In terms of iconicity, this intermediary category is formally marked for both the possession and absolute forms. Therefore, absolute and possession markings also encode conceptual distance between possessor and possessee.

INALIENABILITY, ALIENABILITY, AND MARKEDNESS

As the absolute construction is characterized by the absence of possession, it seems unnatural for inalienable noun references to be conveyed as absolute nouns -- since inalienable nouns with no further derivational morphology tend to be OBLIGATORILY possessed in the language. To translate this into markedness, inalienable nouns are expected to be unmarked possession but marked absolutes. As it happens, the absolute marker -txi formally expresses exactly the markedness property of inalienable noun references. That is, such inalienable references in Apurinã are expressed by overt rather than covert absolute markers because, by default, they can only be used as marked absolutes. The role of the absolute marker -txi is to cancel possessor-possessed intimate/close/bound/inherent relationships which are intrinsic to inalienable references. On the opposite side, if a noun is alienable, it can be OPTIONALLY possessed; therefore, there is nothing unnatural if it occurs non-possessed, which motivates, also by default, the unmarked property of alienable nouns in absolute constructions. That is, there is no intimate/close/bound/inherent relationships intrinsic to alienable nouns and, thus, if no such specific relations are being 'broken,' no need to 'signal' (formally) nominal associations in which the possessed noun is (typically) alienable. Finally, the nouns in the intermediary category may be considered marked for both absolute and possessed forms.

CONCLUSION

In conclusion, besides distinguishing attributive from possession constructions, possession marking interacts with absolute marking in Apurinã to form three marking patterns that grammatically establish three noun categories. The nouns may have been originally classified as belonging to one of the three classes based on their functional roles -- which may have been socially or culturally bound. These roles characterize a noun as being in a more alienable, intermediary, or more inalienable relationship. The semantics behind the system can be partially identified when the two extremes of the alienability-inalienability continuum are contrasted, in which prototypical alienable and inalienable nouns are in the opposite ends. This restriction of recognizable semantic features to the extremes of a continuum suggests that the semantics of the noun categonzation system is bleaching, and a more grammatical distinction is being established. Finally, the patterns of possession and absolute marking encode iconicity and markedness properties intrinsic to alienability or inalienability.

REFERENCES

Carlson, R. and Payne, Doris. 1989. Genitive Classifiers. In Proceedings of the Fourth Meeting of the Pacific Linguistics Conference, eds. R. Carlson, S. DeLancey, S. Gildea, D. Payne, and A. Saxena. University of Oregon, Eugene.

Chappell, H. and McGregor, W. 1989. Alienability, Inalienability and Nominal Classification. BLS15.

Facundes, S. da S. 1994. Noun Categorization in Apurinã (Maipuran, Arawakan). MA. Thesis, University of Oregon.

Nichols, J. 1988. On Alienable and Inalienable Possession. In Shippley (ed.) In Honor of Mary Haas. From The Haas Festival Conference on Native American Linguistics; pp.557-609. Berlin: Mouton de Gruiter.

Payne, David L. 1993. Classifrcation of Maipuran (Arawakan) Languages Based on Shared Lexical Retentions. In D. C. Derbyshire and G. K. Pullum Handbook of Amazonian Languages, vol. 3, pp.355-499.

Seiler, H. 1983. Possession: an Operational Dimension of language. Tübingen: Gunter Narr Verlag Tübingen.