Sie sind auf Seite 1von 6

This is a reply to Achille Varzis paper A Note on the Transitivity of Parthood < http://www.columbia.edu/~av72/recent.html >.

Both will be published in Applied Ontology. An Interdisciplinary Journal of Ontological Analysis and Conceptual Modeling (2006).

What About and - in -Parthood? Reply to Varzi


Ingvar Johansson Institute for Formal Ontology and Medical Information Science, Saarland University, Germany Achille Varzi ends his note [1] on my paper [2] by saying: Parthood is transitive, -parthoodfor many values of is not. Let me call this view the (Simons/Casati/Varzi) predicate modifier account of non-transitive parthood relations [3, 4]. I now agree with it, and as stated in this quotation, I agreed with it already in the paper Varzi (partly) criticizes. My earlier problem was that I thought (mistakenly) that Simons, Casati, and Varzi took it for granted that all -parthood predicates are subsumable under parthood. However, since Varzi now makes it perfectly clear that it should not be so understood (p. 4), there is no longer any reason for me to present my views as some kind of alternative to the predicate modifier account. My and Varzis present disagreements are situated within this account. To my mind, they are centered around two main issues. First, contra Varzi, I cannot see anything misleading or wrong in the proposal I put forward: Seemingly intransitive and non-transitive binary parthood predicates, both in everyday and in scientific language, are in every case hiding a reference to a third relatum (p. 180). Our concern is not classical formal mereology as such, but its relation to parthood predicates in non-formal languages. The formal-mereological parthood relation is a binary relation (we all agree), whereas -parthood relations, both in natural languages and in formal mereology, can (I claim) be more than binary. Second, I take Varzi implicitly to have the view that applied mereology sometimes uses the pure parthood relation and sometimes -parthood relations, whereas I have started to think that applied mereology can work only with -parthood relations. It seems to me that just as similarity in our spatiotemporal world is always similarity in a certain respect, each thisworldly parthood relation is parthood in a certain respect. In particular, it means that I regard spatial parthood as a (transitive) -parthood relation, whereas Varzi seems to regard it as simply a parthood relation [5, Appendix: Outline of a Theory]. I would not be astonished if these present disagreements are signs of a need to develop within the predicate modifier account a detailed theory of what different kinds of predicate modifiers may look like, and why some of them preserve the transitivity of pure parthood and why some of them do not. Perhaps, in parallel with such a theory, even a philosophy of applied mereology analogous the philosophy of applied mathematics has to enter the philosophical world. I will end my reply with a comment on this. Since a theory of predicate modifiers should be able to tell what makes some -parthood predicates transitive (e.g., spatial part), some intransitive (spatial 60%-part), and some non-transitive (large spatial part), it will in all probability have practical implications for informatics. The authors and developers of all domain-specific computer ontologies such as the Gene Ontology (GO) [6] and the Foundational Model of Anatomy (FMA) [7], which contain parthood relations, have better to know (i) whether their -parthood relation(s) are transitive or not, and, (ii) if not, whether these relations can nonetheless for the practical

purposes at hand be regarded as if they are transitive. Therefore, it is nice that our discussion takes place in this journal, Applied Ontology. Below, I will make some more concrete remarks on Varzis note in order to sustain my view that our present disagreements are connected with some, as far as I know, hitherto unexplored logical and mereological problems. 1. There are in logic two seldom used distinctions that I want to highlight: (1) the distinction between true monadic predicates such as is round and relational predicates such as is as round as this circle, and (2) the distinction between true binary relation predicates such as is mother of and relative products of binary relations such as is grandmother of. My four examples can be formalized as follows (note that R is used as a symbol for relations with different arity): (1a) Rx (1b) z Rxz (or Rxa) (2a) Rxy (2b) R/Rxy =df z (Rxz & Rzy) (= x is round) (= x is as round as this circle) (= x is mother of y) (= x is grandmother of y)

What is the arity of R in these four cases? With respect to (1a) and (2a) the answers are straightforward: in (1a) R is a unary predicate and in (2a) R is binary. But what about (1b) and (2b)? In these cases, one can distinguish between at least two kinds of arity for R: arity with respect only to free variables and arity with respect to all variables, respectively. For some unclear reason, Varzi accepts only arity of the first kind, let us call it free arity, whereas in my paper I speak about arity of the second kind, total arity. Using this distinction, we can say that in (1b) R has a unary free arity but a binary total arity, and that in (2b) R has a binary free arity but a ternary total arity. The fact that in (2b) z is a bound variable is what makes it exist only implicitly in the definiendum R/Rxy. If in this definition we insert definite values in the variables (x = a, y = b, and z = c), it becomes quite obvious that R/R involves three relata, even though only two are explicitly mentioned in the definiendum and focussed on when the predicate is used in ordinary conversation; when we speak of a certain grandmother we note that there is a mother, too, but we put her in the background. We get: (2c) R/Rab =df (Rac & Rcb). Varzi writes that the number of arguments of the definiendum [of relative products] is perfectly binary (p. 5-6, italics mine). Even though this in conformity with tradition in logic, it seems in the context of our discussion to be a curious denial of the obvious: relative products allow a distinction between free (and explicit) arity and a total (and partly implicit) arity. And the same goes for relational predicates. 2. In my paper [2], I present 15 examples of parthood predicates where transitivity does not seem to hold. I claim that two of these (6 and 15) lack transitivity because of equivocations, but that all the others lack transitivity because, linguistic appearances notwithstanding, they are not true binary relations. That is, they have an implicit reference to (at least) a third term. Like Varzi, I am firmly convinced that all true binary parthood predicates are transitive. Now, Varzi claims that my third-and-hidden-relatum solution is in all these thirteen cases either wrong or not really important (I am not sure how exactly to interpret him in this respect), and that, instead, transitivity founders because in these cases we have not simple parthood transitivity, Pxy & Pyz Pxz, but (p. 6):

(8)

(Pxy & Fxy) & (Pyz & Fyz)

(Pxz & Fxz)

Obviously, if Fxy is transitive then the -parthood Pxy & Fxy is transitive, too. Varzi is not by means of (8) stating a sufficient condition for failing transitivity. If Fxy = x is spatially smaller than y, then (8) yields the transitive -parthood proper spatial parthood, but if Fxy = x is 60% of y, then (8) yields the intransitive -parthood 60%-parthood; and we ought to be able to say something about what makes these two Fxys differ. In my paper, I claim that in the latter case there is also a third relatum. What then about functional parthood, which Varzi uses as his illustrative example? I can in relation to this -parthood make my point even without using relative products, which Varzi focuses on. In my opinion, which Varzi accepts, Fxy is now an abbreviation for there is a w (e.g., a blood volume) such that x (the heart) makes something happen to w (the blood) that is relevant for xs (the hearts) function in relation to y (the circulatory system), and the implication (8) can be re-written as: (8' (Pxy & w Fxwy) & (Pyz & w Fywz) ) (Pxz & w Fxwz)

This means that we are back in my proposed solution: the failing transitivity is explained by the existence of relations with a total arity of at least three. My solution does not depend on whether or not functional parthood is a primitive ternary relation, a relative product in the traditional sense, or a qualified relative product (as I argued). The important point is the fact that there is a third relatum involved; be it referred to by means of a free variable, a bound variable, or an individual name. 3. Despite the fact that we now explicitly agree that there is no subsumption relation between parthood and -parthood, Varzi takes me to task for having false beliefs about subsumption. According to the subsumption principle, what is a true description of a subsuming class is necessarily also a true description of all the subsumed classes. Varzi wants to revise this principle, and puts forward an example that he thinks shows such a need. He claims that what is true of parenthood is not always true of the subsumed motherhood: from the fact that parenthood can be instantiated by a man it does not follow that motherhood can be instantiated by a man (p. 3). Examples of this kind abound. Here is another one: from the fact that being an animal can be instantiated by a snake it does not follow that being a monkey can be instantiated by a snake. The expression the class C is instantiated by c means that c exists, but existence does not add any descriptive characteristics to C. In a paper (from 1936) with the question-title Is Existence a Predicate?, G. E. Moore famously answered No, it isnt [8]. Neither is is instantiated an ordinary predicate or a property. Contrary to Varzi, I think that the subsumption principle should not be used at all in relation to the quasi-predicate is instantiated. 4. In passing, Varzi mentions the relation between parthood and proper parthood (p. 4), and calls proper a predicate modifer, which seems obvious from a natural language grammar point of view. However, if this is true, then also is proper parthood a -parthood. In order to investigate the last statement, let us see what happens if we move from natural language to formal mereology. If we take proper part as the undefined predicate and write Pxy =df either (x is a proper part of y) or (x = y), then there is no positively symbolized predicate modifier, and there can be none. What can here modify parthood so as to yield proper parthood is only the taking away of the second

disjunct in the definiens. If, on the other hand, we take part as the undefined term, then we get the definition x is a proper part of y =df Pxy & not-(y = x), and we get a complex predicate modifier. First it modifies identity by means of negation, and then it needs also conjunction in order to deliver the predicate proper parthood. What in natural language is a simple modifier becomes in formal-mereological language a complex modifier. A problem Varzi has not noted becomes visible: either proper is not a formalmereological predicate modifer or such modifiers have to be allowed to be complex; and one has to be somewhat careful before one says anything about -parthood in general. As I have said already once: The problem of mereological intransitivity is not a problem in formal mereology as such, it is problem that concerns the relation between formal-mereological parthood and parthood predicates in non-formal languages. 5. I will here take the opportunity to make also another remark on the relation between parthood and proper parthood, even if it only indirectly touches Varzis note. But I am sure that it is relevant for the aims of this journal. I think there is a need to distinguish between axiomatic mereology for pure mereologys sake and axiomatic mereology for applied mereologys sake. Confronted with the formal-logical inter-definability of parthood and proper parthood (see point 4), present-day mereologists (but not Peter Simons in his classic Parts [3]) seem unanimously to opt for part as their primitive undefined predicate. However, at least when it comes to -parthood, I think one should for pragmatic reasons choose proper -parthood (e.g., proper spatial part, proper temporal part, proper anatomical part, and proper functional part) as the undefined term. To take just one of an incredible number of possible examples, we need to say that the heart is a proper part of the circulatory system but never that the heart is either a proper part of or identical with the circulatory system. It is confusing for non-mereologists that the mereological term that corresponds to our everyday and empirical-scientific term part is proper part, not simply part. (As far as I have checked, all part_of occurrences in the GO and in the FMA can mean proper part; and I guess that this is also what the original authors meant.) Furthermore, before there is a clear exposition of what different predicate modifiers of part may look like, it is not even certain that the predicates -parthood and proper -parthood are inter-definable for all s. 6. Varzi says that Generally speaking, a predicate modifier is just another predicate (p. 4), and that a predicate modifier for part is always restricting the extension of part (p. 4). To me, these two seemingly innocent statements raise several theoretical questions. Here they are with some complementing brief remarks: a) What does it mean that a predicate modifier is just another predicate? First, can in principle a predicate modifier for parthood be a predicate of any arity? Varzi exemplifies only with unary and binary modifiers, whereas I have argued that there are ternary ones. Second, can a predicate modifier for part be a predicate that does not apply directly to the individuals that stand in the parthood relation, but to the parthood relation itself? Compare good in good pianist with large in the non-transitive predicate large spatial part. As Varzi himself mentions, when we say that a certain pianist is a good pianist, good does not apply directly to the person in question but to his property of being a pianist. Therefore, the sentence x is a good pianist cannot be rendered as x is good & x is a pianist. Similarly, x is a large spatial part cannot be rendered as x is large & x is spatial & x is a part. But Varzi does not use his pianist example in order to develop any specific remarks on his conception of predicate

modifiers. Instead, he continues to write as if predicate modifiers always are unary or binary and always come as conjuncts. This takes me to the next question. b) In what ways can predicate modifiers be connected to the predicate they modify? Varzi uses explicitly only conjunction, but as I remarked in point 4, he might accept modifiers that involve negation. If some modifiers of part modify in the way good modifies pianist in good pianist, then the formal systems for these -parthoods have to contain formal devices that are not to be found in classical mereology. We cannot simply take it for granted that no -parthood involves (with Varzis expression) a departure from the scope of classical mereology (p. 3). c) Is it meaningful to speak, as Varzi does, of the extension of part? If it is, then the extension of the mereological predicate part (P) is absolutely everything, and Russells paradox knocks on the door; even though no mereologist seems to hear it. Let it be added that Peter Simons has made a brief comment [3, p. 102]. In my opinion, the paradox enters mereology in the following way. The statement formula Pxx is true for every existing entity, since either x is a proper part of x or x is identical with x, and the latter disjunct is true for everything. This means that the extension of the predicate is P contains the x referred to by means of the predicate that gives rise to Russells paradox: is a set of those sets which are not members of themselves. If because of this, in the spirit of Russell, one regards the unqualified expressions all sets and all propositions as syntactical misfits or meaningless expressions, then one should regard all parts in the same way. Most present-day set theorists seem to take Ernst Zermelos qualification as the best solution to Russells paradox as applied to set theory [9]. In my opinion, it is time for extensional mereology to try, so to speak, to find its Zermelo. d) Even if we move to proper part in order to get rid of Russells paradox, philosophical issues pop up. Varzi also claims that: the [proper] parthood relation modelled by classical mereology is truly governed by formal ontological principles, i.e., principles that are metaphysically neutral and domain-independent and, therefore, realized or exemplified by any entities whatsoever (p. 2). I would like to question what Varzi says in this therefore-clause. Perhaps the truth is quite the opposite; perhaps the domain-independent pure proper parthood relation is not spatiotemporally realized everywhere but nowhere. Instead of having everything spatiotemporal except the complete history of the whole universe in its extension, pure proper parthood might lack an extension of its own altogether. Let me explain. Proper part is a predicate in formal ontology, and can, I think, be compared with the numerals in mathematics; pure mathematics might also be called a kind of formal ontology. There has always been a strong intuition in and around mathematics to the effect that pure numbers do not have a spatiotemporal existence. Of course, there can in various times and places be quantities such as 5 bananas, 5 cars, 5 stones, 5 cm, 5 kg, and so on, but there seems to be no spatiotemporal entity that can be referred to by the numeral 5 alone. It seems as if numbers have to be connected either to something that is named by a count noun or to a physical dimension before they can enter the world. On the linguistic level, it looks as if the numeral predicates have to be modified by predicate modifiers such as count nouns and terms for non-formal dimensions before they can refer to anything spatiotemporal. Similarly, it is perhaps the case that proper part lacks spatiotemporal extension, even though specific -parthood predicates such as proper spatial part and proper anatomical part have such extensions. As there is something very special about the pure mathematical numbers, there seems to me to be something equally special about the basic entities of ordinary formal ontology, too.

This was my brief plea for a philosophy of applied mereology. It can be summed up as follows: Perhaps proper -parthood is to proper parthood what quantities are to numbers. If so, since the latter relationship has given rise to a philosophy of applied mathematics (often under the name of philosophy of measurement), the former relationship ought to give rise to a philosophy of applied mereology.

Acknowledgments This reply was written under the auspices of the Wolfgang Paul Program of the Alexander von Humboldt Foundation, the Network of Excellence in Semantic Interoperability and Data Mining in Biomedicine of the European Union, and the project Forms of Life sponsored by the Volkswagen Foundation. I would like to thank Pierre Grenon and Fabian Neuhaus for comments. References [1] [2] [3] [4] [5] [6] [7] [8] [9] Varzi, A. C., 2006, A Note on the Transitivity of Parthood, in this issue of Applied Ontology. Johansson, I., 2004, On the Transitivity of the Parthood Relations, in H. Hochberg and K. Mulligan (eds.), Relations and Predicates, Frankfurt: Ontos Verlag, pp. 161-181. Simons, P. M., 1987, Parts. A Study in Ontology, Oxford: Clarendon. Casati, R. and Varzi, A. C., 1999, Parts and Places: The Structures of Spatial Representation, Cambridge (MA): MIT Press. Casati, R. and Varzi, A. C., 1994, Holes and Other Superficialities, Cambridge (MA): MIT Press. The Gene Ontology: http://www.geneontology.org/ The Foundational Model of Anatomy: http://sig.biostr.washington.edu/projects/fm/AboutFM.html Moore, G. E., 1959, Philosophical Papers, London: George Allen & Unwin. Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/contents.html, entry Russells Paradox (A. D. Irvine, 2003).

Das könnte Ihnen auch gefallen