Locative terms and spatial frames of reference in

Locative terms and spatial frames of reference in Wan1
TATIANA V. NIKITINA
Abstract
The paper is a study of spatial expressions in Wan (Mande, Côte d’Ivoire). The
system of locative postpositions of Wan comprises two semantic classes which
have distinct historical sources: terms referring to intrinsic parts of inanimate
objects, and body part terms. I discuss differences in the use and interpretation
of locative terms from these two classes. Some of these differences can be explained by different degrees of grammaticalization, in that body part terms are
to a lesser extent specialized for expression of abstract spatial relations than
object part terms.
I illustrate differences in interpretation between body part postpositions and
object part postpositions by a pair of nearly synonymous postpositions tā ‘on
top of’ and pēN ‘at the head of’, which rely in their interpretation on different
frames of reference (fixed armatures vs. object-centered, Levinson 2003: 77–
79). My analysis is based on the notion of internal asymmetry of the reference
object, which allows the body part postpositions to pick out a spatial region
in a way that is independent of the object’s actual orientation. This example
illustrates one way in which the internal asymmetry of the reference object
constrains the interpretation of spatial adpositions.
1. This study is based on the author’s fieldwork in Kounahiri and Abidjan in 2001–06, and is
part of a language documentation project supported by the Swiss National Foundation (SUBJ
062156.00). All examples come from the Myanmu dialect; I could not identify any significant differences between the two dialects of Wan (Myanmu and Kenmu) in the expression of
spatial meanings. For discussion and comments on an earlier draft of the paper I thank Beth
Levin, Valentin Vydrine, Brook Danielle Lillehaugen, the editors and the two anonymous referees, as well as the audience of the Colloque annuel de l’Institut de Linguistique Appliquée
(“Langues ivoiriennes – La théorie et la pratique”), which took place in Abidjan on February
15–16, 2006.
JALL 29 (2008), 29–47
DOI 10.1515/JALL.2008.002
0167–6164/08/029-0029
©Walter de Gruyter
30
Tatiana V. Nikitina
This paper is a study of terms used for spatial reference in Wan, a Southeastern Mande language spoken in central Côte d’Ivoire (Ivory Coast). The system
of locative terms found in Wan is interesting for the typology of spatial expressions in that it comprises two separate semantic classes of lexical items.
Locative terms referring to intrinsic parts of inanimate objects co-exist in the
language with terms that originally referred to body parts of animate objects.
Members of both classes can function as nouns designating a specific part of an
object; some locative terms can also be used as postpositions, which describe
a specific spatial relation with respect to a reference object, i.e., localize the
Figure in relation to a particular search domain.
The paper focuses on the postpositional use of the locative terms. In Section 1, I outline the basic typological characteristics of Wan locative postpositions. Section 2 contains an overview of the semantics of specific postpositions from the two subclasses of locative terms, as well as a brief discussion
of the role of context in the conceptualization of a specific spatial relation. In
particular, I analyze one example of variable adpositional marking, where the
pragmatic relevance of a certain aspect of the spatial relation is the determinant
factor in the selection of postposition.
In Section 3, I describe differences in the use of locative terms from the two
subclasses. I argue that they can be explained by a difference in their degree of
grammaticalization, so that body part terms are to a lesser extent specialized
for expression of abstract spatial relations than object part terms.
Finally, in Section 4 I discuss a contrast in the interpretation of two nearly
synonymous members of the two classes (tā ‘top’ vs. pēN ‘head’). I show that
their interpretation as postpositions relies on different linguistic frames of reference (fixed armatures vs. object-centered, using the terms of Levinson 2003:
77–79) and relate this difference to a contrast in functional asymmetry between
the types of reference objects used with the two subclasses of locative terms.
In particular, objects combining with body part postpositions tend to have a
well-defined inherent asymmetry, while the use of object part postpositions is
not restricted to asymmetric objects. The co-existence of the two types of spatial terms in one language makes it possible to investigate the role of inherent
asymmetry in the interpretation of spatial adpositions.
1. Typological characteristics of locative postpositions
In Wan, like in a typical Mande language, the head noun follows the possessor,
and the basic word order can be characterized as S-O-V-X: the object always
precedes the transitive verb, and all oblique arguments follow it, marked by
postpositions. Position with respect to the finite verb can be used in Wan as the
basic syntactic test for distinguishing between noun phrases and postpositional
Locative terms and spatial frames of reference in Wan
31
phrases, since some of the postpositions, as will be discussed later, are formally
indistinguishable from the locative nouns from which they are historically derived. Postpositional phrases, but not noun phrases, commonly occur after the
finite verb; postpositions also differ from nouns in that they cannot combine
with a definite marker or modifiers such as adjectives.
Wan postpositions can be divided into two major classes based on their use
and their historical source: locative postpositions and functional postpositions.
Like in many languages of the world, locative postpositions in Wan are derived
from relational nouns with locative meaning; this is the class of postpositions
discussed in this paper. It is characteristic of the locative postpositions that
they can be used predicatively with a copula, characterizing the position of the
subject with respect to a reference object, as in (1):
(1)
n`
ã`
ã
kú
wā2
1sg+cop house under
‘I am in the house.’ (literally, “under house”)
Functional postpositions differ from the locative ones in that they have no
corresponding nouns. Functional postpositions are typically used to mark a
specific thematic relation between a verb and its oblique argument; the set of
functional postpositions includes markers of the instrumental/comitative role
(yā ‘with’), the role of addressee (lÈN ‘to’), the role of recipient (m´ã ‘to’, used
only with the verb kÉ ‘give’). Consequently, they cannot be used predicatively
with a copula and are not involved in the expression of spatial relations. I will
not discuss the class of functional postpositions further in this paper.
In the discussion that follows, I assume that a typical spatial adposition localizes an object or an event (the Figure) with respect to a reference object (or
the Ground). In (2), where the position of a dog is defined with respect to a
house, the house functions as the reference point of the spatial relation in.
(2)
The dog is in the house
The lexical meaning of a spatial adposition typically includes information about
the search domain, or the region in space that the adposition picks out as containing the Figure. In (2), the search domain is the inside part of the house; this
information is specified by the preposition in.
The relation between the Figure and the search domain can be static (as in
The paper is lying on the table) or dynamic, and in the case of a dynamic
2. Wan is a tonal language. The tones are marked as follows: á – high, ā – mid, à – low. The
following abbreviations are used in the interlinear gloss: cop – copula, def – definite marker,
incl – inclusive 1st person, neg – marker of negation, past – marker of past tense, pl –
marker of plurality, prgr – marker of the progressive aspect, rfl – reflexive pronoun.
32
Tatiana V. Nikitina
relation, the search domain can be the source, the path or the goal of motion
(cf. The dog runs from/through/into the house). Indo-European languages often express both the search domain and its relation to the Figure with a single
lexical item, usually a preposition. In English, prepositions in, into, from and
through can pick out the same search domain (the inside part of the reference
object) but differ in terms of its encoded relation to the Figure: in the house is
used when the relation is static, into the house characterizes the search domain
as the goal of motion, from the house and through the house refer to the search
domain as the source and the path of motion respectively. Heine, Claudi and
Hünnemeyer (1991: 144) describe this type of spatial adposition as portmanteau markers combining “the functions of both V- and N-adpositions”.
Wan, as well as many other Mande languages, lack any portmanteau adpositions of this kind. Not only are goals of motion expressed by the same
postposition as static locations (absence of adpositions specialized for the expression of goal is rather widespread cross-linguistically), but Wan also lacks
any postposition specialized for the expression of source of motion. The two
types of information – the search domain and its relation to the Figure – are
encoded in this type of language by entirely distinct lexical classes. Only the
search domain is specified by locative postpositions, while its relation to the
Figure is either encoded in the predicate of the sentence or inferred from context (for further discussion of the absence of portmanteau adpositions in Wan
and the source vs. goal ambiguity, see Nikitina forthcoming).
Another characteristic semantic feature of Wan postpositions is neutralization of the distinction between internal parts of the reference object and external regions of space adjacent to those parts. The search domain picked out by a
Wan postposition is never restricted to an intrinsic part of the object, but always
covers the adjoining region. As a result, no distinction can be made between
localization in contact with the reference object vs. localization in the object’s
proximity. This is illustrated by a list of postpositions with their basic lexical
meanings in (3).
(3)
gw`
ã gó
gw`
ã tā
gw`
ã mì
gw`
ã wā
‘in the calabash’, ‘on the inner surface of the calabash’
‘above the calabash’, ‘on top of the calabash’
‘near the calabash’, ‘on the outer surface of the calabash’
‘under the calabash’, ‘on the bottom of the calabash’
In English, the lack of distinction between the intrinsic part and the external
region characterizes only a small set of prepositions, such as over or under,
which do not necessarily specify whether the Figure and the reference object
are in contact or the Figure is located in the area of space detached from the
reference object. In Wan, this syncretism is characteristic of the system of locative postpositions in general; no postposition is associated exclusively with an
external part (cf. English above), or with an internal part of the reference object
Locative terms and spatial frames of reference in Wan
33
(cf. English on). This syncretism is a common feature of locative adpositions
in other Mande languages, as well as in some Niger-Congo languages outside
of the Mande family.3
2. Two lexical sources of spatial postpositions
Like in many other Mande languages, locative postpositions of Wan fall into
two classes based on the semantic type of relational noun they are derived
from. One class is derived from nouns referring to parts of inanimate objects
(I will refer to them as object part postpositions). The other class is derived
from nouns referring to body parts of people and animals (body part postpositions). The distinction between the two classes of postpositions is based on
the postposition’s lexical source, as determined by the semantics of the corresponding noun (nouns that refer to parts of inanimate objects vs. nouns that
refer to human body parts). In this section, I introduce individual members of
the two classes; differences in the use of the two classes of postpositions and
their corresponding nouns are discussed in Section 3.
2.1. Inventory of the object part postpositions
The shape of the Figure is generally ignored in the selection of a spatial postposition in Wan, whereas the shape of the reference object might be important.
Thus, the postposition lé ‘on the surface of’, ‘at the edge of’, typically specifies a relation with respect to a two-dimensional object. Its corresponding noun
refers to the surface or edge of an entity conceptualized as “flat” (e.g., a piece
of cloth). In postpositional phrases with lé ‘on the surface of’, ‘at the edge of’,
the surface can be oriented vertically, as in ´ã lé ‘on the face’, or horizontally,
as in zī é lé4 ‘on the road’ or bā lé ‘in the field’, but a general restriction on
the use of the postposition lé appears to be the uniqueness of the surface of the
reference object (i.e., the object is typically flat).5
3. Ameka (1995: 149) discusses the contrast between the semantics of internal part and that
of a region in Ewe, where the two meanings are sometimes distinguished by the use of a
possessive connective. For a discussion of a semantic extension from internal part to the space
projecting from that part in Mixtecan (Oto-Manguean family) see Hollenbach (1995: 171–
172) and references therein.
4. Here, and in a number of examples to follow, é is the definite marker; for its use in a full
sentence, see Example (4).
5. Outside of the purely spatial domain, the postposition lé is used to mark the progressive aspect
on verbs (there is an example of this construction in (4), which can be literally translated as ‘I
am at walking’). This postposition is also a component of the complex possessive postposition
ÒN-lé ‘in one’s possession’ (derived from ‘in hand’).
34
Tatiana V. Nikitina
Localization with respect to a three-dimensional object is described by several postpositions that pick out different parts of the Ground as the search domain for the spatial relation. Wan distinguishes between four major intrinsic
parts of a three-dimensional object: the top, the sides, the inside, and the bottom. Roughly speaking, this distinction is based on the relative position of the
intrinsic part with respect to the gravitational axis (top, middle, or bottom) and
on the notion of containment (inside vs. outside).
Containment is expressed by the postposition gó ‘in(side)’, derived from the
noun gó ‘inside’, which refers to the internal part of an object: yrĒ é gó ‘in/on
the tree’6 , yí é gó ‘in the water’.
Localization in the upper part of the reference object, on its upper surface or
above it is expressed by the postposition tā ‘on, above’: kpē é tā ‘on the chair’,
yí é tā ‘at the surface of the water’, ‘above the water’. The corresponding noun
is tā ‘surface, upper part’.
Localization in the bottom part, on the bottom surface or below the reference
object is expressed by the postposition wā ‘under, at the bottom of’: yrĒ é wā
‘at the bottom of the tree’, ‘under the tree’, kú wā ‘in the house’ (literally,
‘under house’, i.e., in the bottom part of the area delimited by a hut; when used
as a postpositional phrase, this form is often contracted into kúā ‘at home’).
The corresponding noun is wā ‘bottom’.
Finally, localization on the sides of the object and in the adjacent external
regions is expressed by the postposition mì ‘by, near, adjacent to’: yrĒ é mì
‘by/near a tree’, Ǹ Ō mì ‘on my hand’ (e.g., about something that got stuck to
the hand). The corresponding noun is mì ‘side’.
There is an additional type of localization, expressed by the postposition
trāgá which refers to the space in the middle of the Ground: kŌN é trāgá ‘in
the middle of the village’. Like the rest of the postpositions from this class,
trāgá can also refer to localization in the region external to the reference object.
Thus, when a group of objects serves as the Ground, trāgá is commonly used
to locate the Figure in the space between the objects: kŌN pīlŌN trāgá ‘between
two villages’. The corresponding noun refers to the middle of an object.
There are a few locative postpositions in Wan that do not pick out any specific region of the reference object as their search domain. Thus, the postposition n`
Ẽ ‘at the place of’, derived from the noun n`Ẽ ‘place’, refers to a place of
action as a whole and typically combines with abstract nouns whose referents
cannot be divided into internal spatial regions, such as nominalizations referring to events (‘in the place of X’, where X denotes an event). Another locative
postposition, plá ‘at one’s place’, is used exclusively with animate reference
objects and typically localizes the Figure at an individual’s “home”, i.e., picks
6. In Wan, trees are conceptualized as containers and the foliage is described as the “inside” of
the tree.
Locative terms and spatial frames of reference in Wan
35
out an abstract region defined by the relation of possession, or the relation of
control, with respect to an individual.7 Unlike the postpositions described earlier in this section, neither n`
Ẽ ‘at the place of’ nor plá ‘at one’s place’ picks out
a specific spatial region of the reference object.
2.2. Inventory of the body part postpositions
Postpositions derived from body part terms are most often used for locating
the Figure with respect to an animate Ground – an object that naturally has a
“head”, “back”, or “belly”. This is, however, only a general preference. In many
cases, body part postpositions can be used with inanimate reference objects,
when there is a physical or a functional similarity between the relevant part of
the reference object and a body part of a human, cf. gÒN é pēN ‘on top of the
hill’ (literally, “at the hill’s head”), kú é káò ‘in the room’ (literally, “in the
belly of the house”).
Like in the case of the object part postpositions, regions picked out by body
part postpositions can be characterized in terms of their localization with respect to the gravitational axis. Body part postpositions, however, distinguish
between two, rather than three, spatial regions: “top” and “body”; no postposition is derived from the noun referring to feet (c´Ẽ ‘foot’). An exhaustive list of
both object part and body part postpositions can be found in Figure 1.
Localization in, on, or above the top part of the reference object can be expressed by the postposition pēN ‘on top of, above’, derived from the noun pēN,
which refers to the top of the head: yrĒ gb¯Ì̃N é pēN ‘top of the tree’ (literally, “head of the tree’s trunk”). This postposition is roughly synonymous with
the object part postposition tā ‘on, above’ (on differences between the two see
Section 4).
Localization with respect to the non-top part (the “body” of the reference
object) can be expressed by several postpositions depending on other properties. The relation of containment is expressed by the postposition káò ‘inside’,
derived from the noun káò ‘belly’, cf. kŌN káò ‘in the city’.
Localization in the region external to the reference point is expressed by
the postposition dìN ‘near, close to, next to’. This postposition appears to be
derived from the noun dìN ‘hip/side’. Example (4) illustrates the use of dìN as
a postposition.
7. The postposition plá ‘at one’s place’ is found in Wan toponyms referring to villages and other
settlements. Such toponyms typically have the form “ProperName + plá” or “EthnicName +
plá”.
36
(4)
Tatiana V. Nikitina
n`
ã`
ã
táò lé
kú
é
dìN
1sg+cop walk prgr house def close.to
‘I am walking next to the house.’
As I already mentioned in Section 2.2, postpositions derived from object part
nouns define the localization of the Figure based on containment and the position of the Figure in terms of the gravitational axis. Neither of these notions
takes into account the position of the observer. In this respect, the spatial relations that are expressed in Wan by the set of object part postpositions are quite
different from the relations described by the English prepositions behind or in
front of. Thus, localizing the Figure in front of a tree generally implies that the
Figure is located between the tree and the observer or the deictic center; behind
the tree implies that the tree is located between the observer and the Figure.
In Wan, all object part postpositions locate the Figure in terms of spatial orientation of the reference object alone (Levinson’s 2003: 41–43 intrinsic frame
of reference). Some body part terms, however, can express spatial orientation
relative to the position of the observer (the relative frame of reference, Levinson 2003: 43–47). The postposition kúlà ‘at the back of’, derived from the
noun kúlà ‘back’, picks out an intrinsic part of the reference object if that
object has a front/back asymmetry. On the other hand, it can signal that the
reference object is located between the observer and the Figure, as in (5):
(5)
kālĒ áálè dō á
Gōlí gàN
bā é
kúlà
forest big one cop Goli sorghum field def back
‘There is a big forest behind Goli’s plantation of sorghum.’
In (5), the plantation itself is symmetric and not conceptualized as having an
inherent “front” or “back”. Rather, the region referred to as “back” is defined by
the position of the observer or the deictic center. Unless the context introduces
the position of a specific observer (e.g., the protagonist of the narrative), (5)
is interpreted as localizing the forest in the area lying behind Goli’s plantation
from the point of view of the village (here, without additional context, the
village serves as the deictic center, since it is the place from which humans
typically approach plantations).
When the reference object has its own internal front/back asymmetry, the
postpositional phrase formed by kúlà ‘at the back of’ may be ambiguous, i.e.,
the postposition can refer to either of the two different search domains (one
determined by the object’s internal asymmetry, the other by the position of
an external observer). The postpositional phrase in (6) may refer either to an
intrinsic part of the object (the back of the car, as defined by the direction of
the car’s motion) or to a region defined in relation to the observer (in (6), the
speaker):
Locative terms and spatial frames of reference in Wan
(6)
37
n`
ã`
ã
yá lé
gòN é
kúlà
1sg+cop sit prgr car def back
‘I will sit at the back of the car.’ or ‘I will sit behind the car.’
The two possible readings of this sentence derive from two different types of
front/back asymmetry, which are in conflict in this example. On the one hand,
the reference object has its own internal functional asymmetry (the “front” and
the “back” of the car are fixed with respect to the direction of the car’s motion,
as well as the direction faced by the people in the car). On the other hand,
the “front” and the “back” may be defined based on the relation between the
reference object and the external observer. This type of asymmetry emerges, in
Svorou’s (1993: 96) terms, not from the spatial configuration of the object, but
“from the situational interaction with the observer.”
In spite of the general lack of distinction between intrinsic parts, surfaces and
adjacent regions, which is typical of locative postpositions in Wan (see Section
1), the relative frame of reference appears to be incompatible with the internal
part interpretation for objects with an internal front/back asymmetry. Although
(6) can mean ‘I will sit in the area adjacent to the part of the car that is furthest
from the deictic center’ (external region, relative frame of reference), or it can
mean ‘I will sit in the back part of the car’ (internal part, intrinsic frame of
reference), it cannot be interpreted as ‘I will sit in the car on the side that is
furthest from the deictic center’ (*internal part, relative frame of reference).
On the other hand, “back of the tree”, where the tree lacks an internal front/
back asymmetry, might refer to the internal part of the tree that is opposite to
the one facing the speaker, not only to the area adjacent to that part. This suggests that the incompatibility of the relative frame of reference and reference to
an internal part in the case of a car is due to a conflict between the observer’s
perspective, which imposes a “situational” front/back asymmetry, and the object’s internal configuration, which predetermines which part will be counted
as the “back” based on the internal properties of the object.
Similarly, when the speaker is facing the back part of the car, the postpositional phrase in (6) can only refer to either the back seat or the external area
adjacent to the front part of the car. It cannot be interpreted as picking out the
external area adjacent to the functional back of the car, since that area is the
“front” from the point of view of the speaker. The fact that this reading is not
available (*external region, relative frame of reference) can be explained by the
same conflict between the internal asymmetry of the object and the perspective
of an outside observer.
Although kúlà ‘at the back of’ appears to be the only locative term that can
be used as a postposition to express localization relative to the position of the
observer, other body part terms, which only function as nouns, can combine
with object part postpositions to express similar spatial relations. Thus, the
38
Tatiana V. Nikitina
noun áēlí ‘front’ can be used in a postpositional phrase with the postposition
lé ‘on the surface’: kú é áēlí lé ‘in front of the house’ (literally, “on the surface
of the house’s front”).
As an alternative to both the relative and the intrinsic systems of spatial reference, Levinson (2003: 47–50) discusses the absolute frame, where the Figure
is localized in relation to the Ground with respect to a fixed landmark (e.g., The
tree is uphill from the house), or within a fixed system of abstract relations
(e.g., The tree is to the south from the house). It is possible to resort to the
absolute frame of reference in Wan, although speakers only use it occasionally
and it is not central in the linguistic expression of spatial relations.
Cardinal direction terms are derived from body part terms that can only be
used as nouns, via a combination of a body part relational noun and an object
part postposition. Thus, m´
Ẽ lé ‘on the head’ and c´Ẽ wā ‘under the foot’ refer
to the north and to the south, respectively; gbù tā ‘on the left side’ and wlĒ tā
‘on the right side’ refer to the west and to the east (all four combinations can
function both as noun phrases and as postpositions). As discussed in Brown
(1983), it is quite common for languages, including West African languages
(e.g., Mende, Yoruba, Dyola), to associate north with “upwardness” and south
with “downwardness”. Brown (1983) suggests that this association can be explained by the “ubiquitous aligning of north with the top of maps” (135), i.e.,
by the influence of Western school education (see also Heine 1997: 52–57). It
is also possible that the interpretation of north and south as the “head” and the
“feet” respectively is based in Wan on the notions “up the river” and “down
the river”, as the major river in the Wan area flows from the north to the south,
and the expression tÓ m´
Ẽ, literally “river’s head”, is used to refer to a river’s
source. Similarly, when one is facing up the river, the west is located to the left
and the east to the right.
It is interesting that although the set of terms used for absolute reference
appears to refer to abstract relations, what they in fact specify, at least in some
parts of the Wan-speaking area, is localization relative to a local landmark,
the river, which is located to the east of most villages; for such speakers, “to
the east”/“on the right” actually means “toward the river” or “toward the big
water”.8
Besides the set of expressions based on the nouns for “right side” and “left
side”, Wan has an alternative system of reference to east and west. This alternative pair of terms identifies east and west through reference to the rising of
Ẽ ‘place where the sun comes out’) and its setting (yĒté tá
the sun (yĒté ml´
ã n`
8. One of my language consultants moved from his native village, located to the west of a big
river, to a city located to the north of the ocean. He was taught at school that the sun rises in
the east (for him, in the direction of the “big water”), and was surprised to see that in that city,
the sun rose “in the north” instead.
Locative terms and spatial frames of reference in Wan
39
n`
Ẽ ‘place where the sun sets’). The factors that influence the choice between
the two systems of reference are unclear.
2.3. Pragmatic salience and variation in the postpositional marking
My discussion of the semantics of postpositions focused on purely geometrical
characteristics of the spatial region picked out as the search domain, as well
as on the notion of internal asymmetry (geometrical or functional) of the reference object. In this section, I would like to discuss one example where two
different postpositions may be used to refer to the same search domain with
respect to the identical reference object. This variation cannot be explained by
the physical properties of the reference object alone (they remain constant),
and it is also problematic for an account based on the prototypical function of
the reference object (the prototypical function does not change).9 Instead, the
choice of the postposition appears to depend on considerations of pragmatic
relevance, or difference in the construal of the spatial relation as determined
by context.
Consider the following examples, where the same kind of geometrical relation is expressed by two different object part postpositions:
(7)
(8)
DèlÒtÓ gā
bā é
lé
Deloto went field def on
‘Deloto went to the field.’
LÈmÈ á
áólì é
bìlà lé
éé bā é
gó
Leme cop goat def pull prgr rfl field def in
‘Leme is pulling the goat from the field.’
In both examples, the search domain specified by the postpositional phrase is
the area of the field. Although the search domain serves as the goal of motion
in (7) but as the source of motion in (8), this information is not encoded in the
postpositional phrase; rather, it is contributed by the verbs of motion (gà ‘go’
and bìlà ‘pull’), which select for a goal phrase and a source phrase, respectively
(see Nikitina forthc. for details).
Thus, the difference in the postpositional marking is not explained by a difference in the spatial relation described or in the verb selecting the postpositional phrase. Rather, it seems to be due to a difference in the relevant aspect
of the same spatial relation. In (7), the field is referred to as a working site, or
as a portion of land meant to be cultivated, i.e., the reference object is viewed
9. On the importance of the function of the reference object in the selection of spatial adpositions, see Vandeloise (1991, 1994), showing that for objects with the same topological properties the choice of the postposition may vary depending on the object’s conventional use.
40
Tatiana V. Nikitina
from the perspective of its canonical function. The localization in the field is
marked by the postposition lé ‘on the surface of’; this postposition picks out
the region at the surface of a two-dimensional object, and it is the postposition
most frequently used with the noun bā ‘field’.
In contrast, in (8), the salient relation is that of containment: the field is
conceptualized as “containing” the goat, and it is this relation that Leme is
trying to do away with by pulling the animal from the field. Thus, the notion
of containment is more relevant in (8), where the field is viewed from the point
of view of its limits, than in (7), where it is viewed as a surface of cultivated
land, and so the “canonical” postposition for the field is used.10 With animal
Figures, there appears to be a general tendency to refer to the space delimited
by a field with the postposition gó ‘inside’.
It is the contextual construal of the spatial relation in the specific context
that appears to determine the speaker’s choice of a postposition in cases of
such variation (variation in the expression of the same spatial relation is a phenomenon found in many languages, cf. examples from English and Danish
discussed in Sinha and Kuteva (1995: 174–177)).11 As the example discussed
in this section suggests, embedding a locative expression in a specific context
can complicate the assignment of labels to spatial relations. Investigation of
specific contextual effects on the choice of a postposition lies outside of the
scope of the present paper.
In the following sections I return to the problem of co-existence of two
classes of locative terms in Wan and outline basic differences in the use and
interpretation of their members. Figure 1 summarizes the meanings of locative
postpositions that have been discussed in Section 2. The expressions in parentheses are combinations of a body part noun with a locative postposition; in
particular, the body part terms áēlí ‘front’ and c´Ẽ ‘foot’ cannot be used on their
own as a locative postposition. Such combinations can be analyzed as complex
locative postpositions.
10. This contrast between the use of the “standard” postposition lé ‘on the surface of’ that typically combines with the noun and the marked option of using gó ‘inside’ is somewhat similar
to the optionality of Ewe postpositions when referring to involvement in an activity associated with a place, rather than to a physical location within that place, as described in Ameka
(1995: 164–165).
11. In Leech (1969: 162–163) and Fillmore (1975: 17–18), the particular kind of variation between in and on, as in in the grass vs. on the grass, is explained by a difference in perceived
dimensionality of the Ground.
Locative terms and spatial frames of reference in Wan
tā
pēN
kúlà
‘at the back of’
41
‘on top of’
‘on the head of’
gó ‘inside’
káò ‘in the belly’
(áēlí lé ‘in front of’)
mì ‘at the side of’
dìN ‘near, close to’
wā
‘at the bottom of’
(cẼ´ wā ‘at the feet of’)
Additional terms
lé
‘at the surface of’, ‘at the edge of’
trāgá ‘in the middle of’, ‘between’
n`
Ẽ
‘at the place of’
plá
‘at somebody’s place’
Figure 1. Spatial regions associated with major object part and body part postpositions
in Wan
3. Difference in the degree of grammaticalization
The two semantic classes of postpositions differ in their frequency and in the
conditions on their use. The object part postpositions are extremely common.
They impose virtually no restrictions on their object, and have a wide range of
non-spatial uses, cf. temporal uses in trē gó ‘at night’ (with the postposition
‘inside’) or lā gà é lé ‘before your departure’ (literally, “at the edge of your
departure”). Some verbs select for an object part postposition, e.g., the verb pā
‘be capable’ selects for the postposition lé (9). In such uses, the postposition
serves only as a marker of an oblique argument of the verb and carries no
spatial meaning (as reflected in the gloss):
(9)
è
pā
ú
wò
lé Ó
3sg capable wine make le neg
‘He cannot make wine.’
In contrast, postpositions from the body part class are relatively infrequent and
are restricted in their meaning to the expression of spatial relations proper.
They are hardly ever selected for by a verb as markers of the oblique argument.
Besides, even in their locative use the body part postpositions do not combine
with their object as freely as the object part postpositions. Animate objects
42
Tatiana V. Nikitina
form the central class of nouns that occur with body part postpositions. Besides
animate nouns, the use of the body part postpositions is extended to inanimate
objects with internal asymmetry, such as dwellings (a hut, but also a village
or a city), which are conceptualized as having “bellies” (designed to contain
humans), “heads” and “sides”, or mountains and trees, which typically have a
fixed rounded top that resembles a human head.
Such differences in the use of the two classes of locative terms as postpositions can be seen to reflect a difference in the degree of their grammaticalization. Members of both classes (body part terms and object part terms) can be
used as nouns, or one-place predicates: part(Object). In this use, they refer to
a specific part of an object (expressed as the possessor), cf. bl`ãN é tā ‘surface
of the lake’, kú é pēN ‘roof of the house’, à áēlí ‘its front’, à c´Ẽ ‘his foot’.
At the same time, the two classes differ in the scope of their use as postpositions, or two-place predicates marking a type of relation between the reference
object and the Figure: relation-part(Figure, Object). All members of the object part class, but only some members of the body part class, can be used in
this function. Terms for body parts that cannot express the relation per se can
combine with an object part postposition: gb¯Ì̃N c´Ẽ wā ‘at the bottom of the
tree’ (literally, “under the trunk’s foot”) or kú áēlí lé ‘in front of the house’
(“at the surface of the house’s face”). One can hypothesize that as soon as such
body part terms become grammaticalized as postpositions, or two-place predicates expressing a specific relation, they will no longer require the use of the
additional element from the object part class.12
Postpositional uses are not the only piece of evidence that points to a difference in grammaticalization between body part terms and object part terms. The
two classes exhibit different frequency of use as prefixes on transitive verbs. In
Wan, and in Mande more generally, locative nouns can grammaticalize as verbal prefixes specifying the intrinsic part of the object that is directly affected
by the action, cf.:
(10)
kó
tā-kÓ
pīlŌN
1dual.incl upper.surface-cut two
‘Let’s cut it into two parts.’
Verb prefixes like the one in (10) functioned originally as heads of object noun
phrases (since the object precedes the verb, tā in (10) can also be analyzed as
the object: ‘Let’s cut the upper surface into two parts.’). At the current stage,
with some verbs such prefixes can no longer be separated from the verb root
12. In fact, some body part terms can be used postpositionally with their literal meaning, as in The
bone got stuck in his throat, but not with an abstract spatial reading detached from reference
to a specific body part. I consider such literal uses to be instances of a construction different
from locative expressions proper.
Locative terms and spatial frames of reference in Wan
43
(e.g., gò-m`
Õ ‘understand’, derived from gó ‘inside’ and a transitive verb m`Õ
‘hear’).
Several verbs may differ from each other in the prefix part depending on the
part of the object affected by the action:
(11)
a.
b.
c.
è
gbŌ é
mì-gò
3sg pot def outer.surface-clean.up
‘She is cleaning the pot (the outer surface).’
è
bī
lī
é
wā-gò
lé
3sg past bed def bottom.part-clean.up prgr
‘She was sweeping under the bed.’
plÉtī mù é
gò-gò
plate pl def inside-clean.up
‘Clean the dishes (inside).’
In the above examples, the choice of the verb depends on the spatial region of
the object that is acted upon, or directly affected by cleaning (the non-prefixed
counterpart of the verbs, gò, can be translated as ‘take off’, ‘take away’). This
distinction is marked by morphemes that originate from the same source as
some of the locative postpositions – a closed set of terms for intrinsic parts of
inanimate objects.
In contrast to prefixes derived from object part terms, there are very few
cases of prefixes possibly deriving from a body part noun. The only verbs potentially containing a body part prefix that I could find are verbs with the prefix
klā ‘behind’, such as klā-klÉ ‘chase’ and klā-pálò ‘turn upside down’ (e.g.,
the soil). The postposition corresponding to that prefix has the meaning ‘after’,
‘behind’, but there seems to be no locative noun that would directly correspond to the prefix. It is possible that both the prefix and the postposition are
derived from the body part term kúlà ‘back’, but it could also derive from a
different (possibly related) noun referring to an intrinsic part of an inanimate
object. With this potential exception, however, body part terms appear not to
grammaticalize as verbal prefixes.
In this section I showed that the two classes of locative terms differ in the
scope of their use as postpositions, as well as prefixes on transitive verbs. This
contrast can be described as a difference in the degree of grammaticalization
of the two types of locative terms. Terms for object parts show a higher degree
of grammaticalization as markers of spatial relation than terms for body parts,
which either tend to be used exclusively as nouns or have a relatively restricted
postpositional use.
44
Tatiana V. Nikitina
4. Difference in interpretation of the two classes of postpositions
In the previous section I discussed differences in the use of the two classes
of locative terms. In this section, I will concentrate on the difference in their
interpretation. As discussed in Section 2.2, all members of the object part class
are interpreted based solely on the spatial orientation of the object, without
reference to an observer. Spatial relations that make reference to the point of
view of an observer (such as “in front of” or “behind”) are expressed using
body part terms, either as postpositions (kúlà ‘behind’, ‘at the back’) or as a
noun combined with a postposition (áēlí lé ‘in front’). Other body part terms
(pēN ‘head’, káò ‘belly’) are interpreted in terms of the intrinsic orientation of
the object, just as postpositions derived from object part nouns.
Within the intrinsic frame of reference, Levinson (2003: 77–79) distinguishes
between three possible subtypes. The first subtype imposes on the reference object a fixed armature, oriented gravitationally, and this armature provides the
object with “top”, “bottom” and “sides” in a way that is independent of the
internal asymmetry the reference object might have. For example, the “top” of
a table in such a system will be the part located higher than all other parts of
the table on the gravitational axis. The part of the table referred to as “top” will
vary in this system depending on whether the table is standing in its canonical
position or lying on its side.
In the other two subtypes of the intrinsic frame of reference, intrinsic parts
are assigned to the reference object once and for all, based either on the internal axial geometry or on the canonical orientation and functional use of the
object. Unlike in fixed armatures, in this system the surface referred to as the
“top” of the table will stay the same if the table is rotated. I will refer to both
such systems as object-centered (Levinson 2003: 77–78), making no further
distinction between systems relying on axial geometry and systems relying on
functional criteria. The table below summarizes the major types of linguistic
frames of reference distinguished in Levinson (2003: 76–92).
Locative terms and spatial frames of reference in Wan
(12)
45
Major frames of reference and their basic characteristics
intrinsic fixed armatures intrinsic parts defined with respect to
the gravitational axis (change with
rotation): above the ball
object-centered intrinsic parts defined based on the
object’s internal asymmetry (do not
change with rotation): by the cat’s head
relative
defined with respect to the position of
the observer: behind the forest, to the
left of the rock
absolute
defined with respect to a local
landmark or a fixed coordinate system:
uphill, to the north of the village
The two classes of locative postpositions in Wan can be contrasted based on
the subtype of the intrinsic reference system available for their interpretation.
The object part postposition tā ‘on/above’ specifies its search domain depending on the orientation of the object with respect to the gravitational axis, which
may be different from its canonical orientation. The area above a fallen tree
will be described with tā ‘on the upper surface/above’, not with the postposition mì ‘on the side/near’. (The latter would be expected if the association of
postpositions with particular parts of an object were preserved independently
of the object’s actual orientation in space.) In Levinson’s terms, the use of object part postpositions in Wan is based on a system of orientation that operates
with fixed armatures, for the most part ignoring the canonical, or functional,
orientation.
Unlike the object part postposition tā ‘on/above’, the nearly synonymous
body part postposition pēN ‘on top/head of’ characterizes the location of the
Figure in terms of the internal axial geometry of the reference object. The region picked out by pēN ‘on top/head of’ remains constant independently of
whether the tree is in its canonical position, oriented vertically, or is lying on
the ground and is oriented horizontally.
The contrast in availability of different interpretations to the members of the
two classes is summarized in (13):
(13)
Postposition
object part:
body part:
tā ‘on, above’
pēN ‘on top of, above’
Frame of reference
intrinsic – fixed armatures
intrinsic – object-centered
The distribution in (13) can be explained by a difference in the role that internal
asymmetry plays in dividing an object into spatial regions. It is crucial for the
46
Tatiana V. Nikitina
use of the object-centered frame that the “head” of an object is identifiable
independently of the object’s position with respect to the gravitational axis (as
in the case with the fallen tree). At the same time, the object-centered frame of
reference is less likely to be used with objects lacking such internal asymmetry:
the top part of a stone or a ball is more likely to be identified based on its actual
position with respect to the ground, since their canonical orientation is difficult
to determine due to the lack of a well-defined internal geometry or functional
asymmetry.13
In other words, the object-centered interpretation, but not the fixed armatures
one, relies on the internal asymmetry of the object. In Wan, the two classes of
locative terms differ in that body part terms can only be used with animates
or other objects that are functionally asymmetric. This may be the cause of
resistance of body part postpositions to the “fixed armatures” interpretation,
which ignores the object’s internal configuration.
5. Conclusion
In the previous sections, I outlined the basic principles of division of space
in Wan. In particular, I focused on the phenomenon of co-existence of two
classes of locative terms with distinct historical sources – terms referring to
parts of inanimate objects vs. terms referring to body parts. I showed that the
difference in the source of locative terms corresponds to a difference in the
degree of their grammaticalization. Although this paper is only a case study of
spatial expressions in one particular language, the distinction between object
part and body part postpositions is found in other Mande languages, where it
often appears to involve similar differences in grammaticalization.
Finally, I suggested that the difference in the lexical source of two postpositions (tā ‘on top of’, ‘above’ vs. pēN ‘at the head of’, ‘above’) may account
for a contrast in their interpretation. The use of the body part term (pēN ‘at
the head of’, ‘above’) is licensed by a well-defined internal asymmetry of the
object. Due to this internal asymmetry the region conceptualized as the “top”
can be identified in a consistent way independently of the actual orientation of
the object. On the other hand, the use of the object part postposition (tā ‘on top
of’, ‘above’) does not presuppose that the reference object is asymmetric. As
a consequence, it is the actual position of the reference object that determines
which part will be referred to as its top.
A different kind of asymmetry is involved in the operation of the relative
frame of reference. Unlike the object-centered frame, which relies on the in-
13. On importance of functional asymmetries in interpretation of body part terms see Svorou
(1993: 94–95).
Locative terms and spatial frames of reference in Wan
47
ternal properties of the object and its canonical/functional position, the relative
frame of reference is based on the interactional asymmetry as construed from
the point of view of the observer. In Wan, the relative frame of reference is only
available with body part terms.
The correlation between the frame of reference and the object’s properties
demonstrates the importance of the notion of asymmetry in the interpretation
of spatial expressions.
Stanford University
ILI RAN, St. Petersburg
References
Ameka, Felix K. (1995). The linguistic construction of space in Ewe. Cognitive Linguistics 6:
139–181.
Brown, Cecil H. (1983). Where do cardinal direction terms come from? Anthropological Linguistics 25 (2): 121–161.
Fillmore, Charles J. (1975). Santa Cruz Lectures on Deixis: 1971. Bloomington: Indiana University
Linguistics Club.
Heine, Bernd (1997). Cognitive Foundations of Grammar. New York: Oxford University Press.
Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer (1991). Grammaticalization: A Conceptual Framework. Chicago: The University of Chicago Press.
Hollenbach, Barbara E. (1995). Semantic and syntactic extensions of body-part terms in Mixtecan:
The case of ‘face’ and ‘foot’. International Journal of American Linguistics 61 (2): 168–190.
Leech, Geoffrey N. (1969). Towards a Semantic Description of English. Bloomington: Indiana
University Press.
Levinson, Stephen C. (2003). Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge: Cambridge University Press.
Nikitina, Tatiana V. (forthcoming). Subcategorization pattern and lexical meaning of motion verbs:
A study of the Source/Goal ambiguity in Wan. Linguistics.
Sinha, Chris and Tania Kuteva (1995). Distributed spatial semantics. Nordic Journal of Linguistics
18: 167–199.
Svorou, Soteria (1993). The Grammar of Space. Amsterdam/Philadelphia: John Benjamins.
Vandeloise, Claude (1991). Spatial Prepositions: A Case Study in French. Chicago: The University
of Chicago Press.
Vandeloise, Claude (1994). Methodology and analyses of the preposition in. Cognitive Linguistics
5 (2): 157–184.