Professional Documents
Culture Documents
1
The following section derives from but supersedes Chilton (2005).
2
If there is a measurement expression, e.g. John is ten metres in front of the tree, this expression
is absorbed into the spatial relation expression as part of the vector ‘label’, thus: ten metres in
front of.
3.1 Vectors, discourse entities and reference frames 53
tree
in front of
John
Consider first how the two sentences below appear in the abstract deictic
space:
(1) John is in front of the tree
John
in front of
tree
Another way of making the point is to note that in objective reality, and
truth conditionally, all physical objects referred to (here John and the church)
have the same absolute spatial positions, but in the conceived situation, which
is what the DSM represents, the relationships are represented differently. This
is what deictically anchored coordinate systems and position vectors enable
us to model explicitly. The DSM does not of course contain all the infor-
mation needed for a full spatial conceptualisation, since it does not represent
the three spatial dimensions. But as already noted, the conceptualisation that
arises in the course of processing discourse involves many parallel systems,
one of which is the three-dimensional spatial system of representation. What
the DSM does is model the fundamental deictic scaffolding.
Consider another spatial preposition, opposite. Once again, this preposition
requires its own spatial model, but such a model, while it will tell us about the
schematic semantics of this word in English, will not tell us about distancing
and directional effects that emerge in discourse. It is just the latter that DST
tells us about. Specifically, the d-axis is used for the mental distance and
direction that arises in discourse rather than in objective physical space. The
sentences (3) and (4) are denotationally equivalent, but have different con-
ceptual significance.
(3) The pub is opposite the post office
is ‘about’ the post office). Because the d-axis is a scale giving us relative
closeness to S, and because vectors give us directionality, we can represent
discourse foregrounding phenomena by positioning one referent closer to 0 on
the d-axis than another referent. In visual experience the ground is what
appears relatively more distant (backgrounded) in space. In the abstract
DSM these phenomena are of course quasi-spatial (or metaphorically spatial)
and constitute what might be called ‘conceived’ distance. The DSM diagrams
handle this very simply by changing the relative positions of the referents: in
(3), for example, the pub is ‘closer’ (more foregrounded) than the post office
and the reference object is the post office more ‘distant’ (more backgrounded).
The tail of the position vector is at the post office and its head at the pub. The
case is reversed for (4).
This section has outlined how the DSM uses the d-axis to model both
conceptualisations of (reduced) physical space and also indicates how the
model gives rise to the notion of attentional distance. The most obvious use of
such a notion is to integrate foreground–background construals into the
geometric modelling of discourse. It is assumed also that, for English at least,
the linear ordering of NPs in clauses will be reflected in the ordering of
referents on the d-axis, barring grammatical constructions for marked topica-
lisation. In the DSMs, the subject NP of a declarative clause will be treated as
‘closer’ to S than an object NP, for example. We have seen that vectors can be
used to represent the relation of objects to landmarks in three-dimensional
space. In DST they are also treated as representing the direction of activity
from a source to a goal. In terms of thematic roles, this means that the tail of
the vector might, for example, be located at the coordinates of an agent, the
head at the coordinates of a patient. Of course, since the vector is directional,
the orientation can be varied in combination with the relatively proximal or
relative distal location of the referents with respect to S. This in turn means
that it is possible to model the conceptualisations associated with passive–
active constructions, raising constructions and similar constructions in a
unified and natural way. We look at these constructions in these terms in
later sections and chapters.
Some linguists find problematic transitive verbs that express spatial rela-
tions. Langacker (1991: 311) examines sentences such as:
(5) a fence surrounds his property
interpreting vectors in the coordinate system, a fact that may reflect the
somewhat variable meanings associated with these and similar verbs. We
can consider using position vectors: for (5) tail at property and head at fence,
i.e. the position of the fence with respect to property. Similarly, for (6) the tail
of the vector is at explosive device and the head at parcel. But neither of these
solutions seems intuitively satisfactory or complete. The other modelling
possibility is to take literally the grammatical transitivity relation (despite
Langacker’s apparent unease about there being any implied ‘active’ element).
Why not follow through the principles of cognitive semantics and accept that
surround and contain are transitive because experience of these spatial
relationships is indeed associated with force in some sense? In this case
surround and contain can be modelled as force vectors. There is justification
for this in the semantics of the spatial preposition in, to which the verbs
surround and contain are conceptually related. Surrounding and containing
may indeed involve force in various types of sentence and context that easily
spring to mind. It has been argued by several linguists that in cannot be
modelled simply by a geometrical schema, by which they mean a purely
spatial configuration (cf. Herskovits 1986, Vandeloise 1991[1986], Tyler and
Evans 2003, Evans and Tyler 2004b). In fact Zwarts (2010: 194–9) surmounts
this problem by using force vectors.
Further, some such spatial verbs denote relations that are geometrically
symmetrical, although linguistic reversal of word order gives them the feel of
being not symmetrical conceptually, as in (3) and (4). These kinds of sym-
metrical relation verbs appear to have a simpler conceptual representation,
since they can be adequately modelled by means of position vectors and
relative positioning of the entities labelled on the d-axis.
Examples (1) to (6) show how position vectors can be used in the obvious
way – to model physical spatial location, abstracting away from the three
physical dimensions and using the d-axis in the base frame of reference to
model relative ‘distance’ of attentional focus. But in conceptual space, the
positioning of two entities in a reference frame need not be confined to
physical entities.
good humour
in
linguist
The discourse entity closer to S on the d-axis (i.e. linguist) is located at the
more distal entity good humour. The position vector, labelled in, has its tip at
the linguist and its tail at good humour – the vector locates the linguist in a
particular state conceived as a container. The DSM configuration, shown in
Figure 3.2, is the same as in Figures 3.1a and 3.1b, and rather than denoting a
physical location denotes a temporary mental state – a temporary property of
the discourse entity linguist.
Possession predicates also lend themselves to treatment as position
vectors. In some languages possession is expressed by spatial meanings:
French has le livre est à Philippe (literally ‘the book is at Philippe’), and
Russian u menya kniga (literally ‘the book is at me’). An English sentence
such as (8):
(8) John has/owns/possesses the book
can be represented by a DSM configured in such a way that the vector for
have, own, possess, etc. locates book at John – i.e. the vector points from
proximal John to distal book. That is, John is at the forefront of attention, the
book is the secondary focus of attention, and the positioning relation is given
by the direction of the vector (Figure 3.3).
Thus Figure 3.3 is the geometrical model for ‘the book is at John’, the
spatial origin of the quasi-spatial sentence. The verb have has the same
conceptual function as a spatial preposition, say at: it locates on entity at
another more focal entity. A position vector gives the position of point
relative to a reference point. Such a point is often the origin of a coordinate
system. Note, however that position vectors in DST generally relate one point
58 Distance, direction and verbs
book
has
John
to another point that is not the origin of the base reference frame. Potentially
all reference entities are origins of embedded reference frames, especially in
the modelling of constructions that represent other minds: individuals other
than oneself have their own reference frames, i.e. coordinate systems.
Taking account of the fact that languages use quasi-spatial locations to
communicate the conceptualisation of possession relations leads us to a
motivated way of modelling property predications in DST in terms of spatial
locations (cf. Anderson 1971, Lyons 1977: 718–24, Frawley 1992: 229–32).
There are a number of considerations and linguistic facts that further justify
modelling of property predications as quasi-spatial position vectors. Sen-
tences such as (5) can be regarded as predicating a property – the temporary
property of being good-humoured – and the means of expression is clearly
spatial.3 The sentences (9) and (10) illustrate two ways English has for
predicating a property of an entity:
(9) Bill has wisdom
3
There is a difference between being in a good humour and being good humoured in that the
latter can be construed as a permanent property. DST does not discriminate such semantic
differences between properties, nor does it discriminate classes of property concept as, for
example, Dixon (1982) does. This does not affect the central claim being made here: that vector
modelling of property predication in general is justifiable because property predication involves
spatial conceptualisation.
3.1 Vectors, discourse entities and reference frames 59
wisdom/wise
Bill