PARTS AND PLACES
THE STRUCTURES OF SPATIAL REPRESENTATION
ROBERTO CASATI AND ACHILLE VARZI,
BOOK REVIEW
BY MARCO AIELLO
|
Parts and Places The Structures of Spatial
Representation Roberto Casati and Achille
Varzi, Cambridge, MA
Book Review by Marco Aiello
Institute for Logic, Language and Computation,
and Intelligent Sensory and Information Systems
University of Amsterdam Plantage Muidergracht
241018 TV
Amsterdam, The Netherlands
E-mail: aiellom@ieee.org Journal of Logic,
Language and Information
10: 270-273, 2001.270 Cambridge, MA:
The MIT Press, 1999. Price: $35.00 (hardcover),
viii + 238 pages, ISBN0262-03266-X. Information
|
|
Since the birth of philosophy, space has
been a topic of investigation which has posed
puzzles and challenges. Today, spatial representation
is receiving even wider attention. From philosophers
to computer scientists, passing through logicians,
linguists, and cognitive scientists, many
minds with different backgrounds are looking
at ways to formally represent space and then
reason with and about spatial structures.
|
The book under review is the joint effort
of two philosophers who have already shown
their nifty-ness in explaining the nature
of some spatial phenomena in Holes and Other
Superficialities (Casati and Varzi, 1994).
The present book is not the continuation
of Holes, rather it takes a broader perspective,
using holes when necessary to explain spatial
peculiarities, but providing a lot of different
examples too. To put it differently, if one
has enjoyed Holes, one will appreciate Parts
and Places even more, if instead one has
to make a choice between the two books from
Casati and Varzi, it is strongly recommended
to start with Parts and Places.
Already in the choice of the title, Casati
and Varzi make the first smart move: rather
than resortingto the usual, but somehow cryptic,
name of the field they are mainly dealing
– that is Mereotopology– they use two much
simpler synonyms – Parts and Places. (The
etymology of the word mereology comes from
the Greek meros = part, and that of topology
also from Greek topos = common place.)The
word Parts accounts for a fundamental property
of spatial entities: they are are built of
simpler constituents and/or together with
other entities build more complex wholes.
The word Places tells us that dealing with
space is dealing with locations and located
entities.
The book brings mereotopology to a wider
public, by using simple words next to more
technical ones and by providing several examples
from our everyday experience of space next
to formal theories of space. Furthermore,
it provides a vast survey of the field, ranging
from the seminal work of Whitehead to the
most recent publications, which makes it
an excellent starting point for the graduate
student and researcher interested in qualitative
formal representations of space. The reader
interested in implementing spatial related
systems might consider the prevalent philosophical
view of the field to be a weakness. She/hewill
sometimes get lost in the philosophical subtleties
of the more abstract concepts and, at times,
wonder about possible implications of the
conclusions reached by the authors on implementations.
The starting point of the investigation is
the identification of appropriate theories
of space. Of course, different tasks need
different approaches. Geologists may resort
to geometry for “measuring the earth,” physicists
to vector spaces to compute the trajectory
of a satellite and engineers to tensor spaces
for building a skyscraper. What then is appropriate
at both the philosophical and cognitive level?
Mereology and topology seem adequate, going
– when necessary – into geometry. The first
four chapters of the volume are dedicated
to presentation of various theories with
particular emphasison the advantages and
drawbacks of each one of them. The underlying
conjecture of Casati and Varziis that the
best way to go is to start with mereology
and then extend it with topology (and other
theory fragments). The following five chapters
each deeply analyzes one different peculiar
property of space; while the following two
deal with applications to specific space-related
issues, such as events in space and maps.
The volume ends with a chapter of open questions,
notes (all grouped at the end of the volume,
which is a bit unhandy), references and a
well compiled index of terms, names and symbols.
Chapter 2 introduces the reader to the spatial
entities that will accompany her/him through
the book: glasses, tables, chairs, donuts,
and, why not, bikinis. But what are the unifying
elements of these sorts of objects? All of
them are composed of hierarchically organized
parts, made of matter and may be one piece
or not. It is the last distinction which
already provides for a surprise.
Mereology, often presented as a theory of
parts and wholes, is actually not enough
to capture the three main features of spatial
entities. Wholeness is a monadic property,
while parthood is a dyadic one which cannot
account for the former. Perhaps, topology
can do the trick. Topology can indeed be
seen as the theory of connection, so self-connectedness
can be expressed, in addition the notion
of set inclusion could work to account for
parthood. This time, it is a very peculiar
spatial phenomenon, that of holes, to impose
a deeper reflection. Topology cannot account
for superficial holes and it has to resort
tothe complement of an object to describe
the hole, i. e., it has to explicitly reference
the immaterial object. The lesson here is
that one needs a formal mereological theory
together with a topological one, the two
theories should be independent, while mutually
related. To put it in the words of Casati
and Varzi: “Mereology alone is too weak;
topology alone is too strong.” The deficiency
of mereology with respect to expressing wholeness
can only be explained within a formal theory
of mereology, and Chapter 3 provides various
such theories. Actually, after an historical
overview of mereology, the authors present
a full hierarchy. The basic one, called ground
mereology M, has only three simple axioms
for the parthood predicate: reflexivity,
antisymmetry and transitivity. No surprise
here, something is part of itself, if two
items are mutually proper part of each other
then they are the same, and finally, a part
of something which is part of a bigger whole,
is also part of the bigger whole. Extensions
of M go into various directions depending
on which extra axioms are added: supplementation,
extensionality, closure principles, and fusion
axioms all provide for acceptable theories
of parts. The top of the hierarchy is what
the authors call general extensional mereology
GEM, and they also point out that GEM has
some spatial flavor: it behaves, as proven
by Tarski in 1935, like set inclusion.
Chapter 4 is another technical piece of work;
equipped with mereological theories, the
authors present topological extensions, again
in the form of a hierarchy. In such extended
theories – i. e., in mereotopology – one
can define predicates for Tangential Part,
Internal Overlap, Internal Proper Part, just
to name a few. A remark, which also applies
to the previous chapter, is that emphasis
is put on the axiomatization of the theories
and on the differences in expressive power
among them, but little is said or shown about
the underlying models and model theoretic
issues.
Boundaries (Chapter 5) are ineludible when
dealing with spatial objects. When there
is some-thing, then there is a boundary.
It may be fuzzy or not, but it is there for
sure. Or is it not? Boundaries are absolutely
spatial entities, but at the same time they
do not take up place. Furthermore, when an
object is divided into several pieces, which
one gets the boundary? Spatial, but also
temporal, boundaries are a source of puzzles
that have to be dealt with when devising
a theory of space. Theories can treat boundaries
in different ways, including disregarding
them from the start, but an ontological commitment
must be taken. The authors suggest, though,
that boundaryless theories are blind of important
distinctions. In a boundaryless theory, there
is no difference between the notion of continuity
and that of contiguity. Another important
characteristic is the impossibility of having
a boundary without having an object: “boundaries
are ontologically dependent entities.” Another
class of puzzling spatial entities which
parasitically live in spatial objects is
that of holes (Chapter 8).Holes, like boundaries,
exist when there is an object they live in:
there is no way of isolating thehole of the
donut while removing the donut. Holes, as
any other parasite, need to be understood
andclassified. In doing so, the authors can
use topology in most cases, but in some other
cannot (forexample topology is blind for
hallows), whence the need for a separate
theory of holes to integratein a full mereotopological
spatial framework.
different way of thinking is required for
counterparts and spatial essentialism (Chapters
6 and9, respectively), because these are
modal notions. Casati and Varzi show how
a theory that accountsfor counterparts and
potential objects can be kept inside the
bounds of first order logic (by includingpossible
worlds in the domain of quantification),
but a first order theory must be extended
by meansof modal operators for the mereological
essentialism. So far, spatially extended
objects and regions have been used as synonyms,
but actually the firstare entities that occupy
the latter. An object has a region as “address,”
and this address can be of manysorts (permanent
vs. temporary, minimal vs. broad, structured
vs. unstructured). The authors proposea theory
to distinguish objects from the regions they
occupy, a theory of location (Chapter 7).
Detailson the interrelation of a location
predicate with the ones for connection and
parthood are spelled out. The last two chapters,
which go towards applications, are somehow
controversial. Chapter 10 isabout temporal
events and their relation to space, an issue
of sure interest. At times, the assumptionseems
to be explicitly made that time can be treated
with the same tools as space. Probably the
drivingforce here is in the linguistic similarities
between spatial and temporal prepositions,
but still it mayturn out to be a dangerous
line of thinking (to get an impression of
some purely spatial behaviors, seeLemon and
Pratt, 1997). Nevertheless, the integration
of time into a spatial framework brings newinteresting
formalizations into the picture, e. g., those
concerning movement. The following chapteris
also concerned with applications. The chapter
on maps (Chapter 11) is intriguing for a
numberof reasons: first, the application
is an important one (think for example of
geographical informationsystems, global positioning
applications and even wearable computing
devices), second, the mainidea of Casati
and Varzi is quite clever: between the real
world we populate and the paper map wehave
in our hands while traveling there is a semantic
layer: that of formal maps. Formal maps arerelated
to the spatial entities of the real world
by an isomorphism, thus the task of the authors
is todefine a semantics for the formal maps.
The attempt is successful, a problem though
resides in thedifficulties to then take the
next step: from formal maps to real world
ordinary maps. The authorssuggest that “this
is analogous to the project of providing
a formal semantics for formal languages.
It will then be possible to ask whether some
features of this semantics can be used to
describe thesemantic structure of ordinary
maps. The aim is to eventually be in a position
to move from thesemantics of formal maps
to the formal semantics of maps.” There is
an obvious worry here: if onthe one hand,
it is true that an isomorphism can be established
between the space to be representedand a
formal map, it is not at all clear how to
establish an isomorphism (or any other kind
of precisemapping) between formal maps and
ordinary maps. Consider a river, it is a
connected spatial entity, in a formal map
it would be represented as a self-connected
entity, but then, in an ordinary map, itwould
not. In fact, road bridges, train bridges,
and other entities are drawn above the river
and makeit a scattered region in an ordinary
map. Hence, even though the idea that formal
maps exist and thatthe laws of mereotopology
govern them is very interesting in principle,
it does not seem to decreasethe gap between
formal theories of maps and actually implementable
systems. All in all, Parts and Places is
a very well thought and carefully written
book. It is clearly intendedfor the philosopher
and the cognitive scientist, but it is also
recommend to the computer scientist. The
latter is likely to be carried away by Casati
and Varzi’s way of explaining the matter,
first posingquestions (that most often one
would not have thought of) and then answering
them. The volume is rich of mind opening
puzzles that borrow from everyday life embedded
in space and time, which is bound to capture
and hold the reader’s attention throughout
the book.
References
Casati, R. and Varzi, A., 1994, Holes and
Other Superficialities, Cambridge, MA: MIT
Press.
Lemon, O. and Pratt, I., 1997, “Spatial logic and the complexity of diagrammatic
reasoning.” Machine Graphics and Vision 6, 89–108.
Marco Aiello Institute for Logic, Language
and Computation, and Intelligent Sensory
and Information Systems University of Amsterdam
Plantage Muidergracht 241018 TV Amsterdam
The NetherlandsE-mail: aiellom@ieee.org
|