Jump to ContentJump to Main Navigation
The Retreat of ReasonA dilemma in the philosophy of life$

Ingmar Persson

Print publication date: 2005

Print ISBN-13: 9780199276905

Published to Oxford Scholarship Online: February 2006

DOI: 10.1093/0199276900.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 23 October 2019



The Retreat of Reason

Ingmar Persson (Contributor Webpage)

Oxford University Press

Abstract and Keywords

Since it has been found that identity consists in the identity of our bodies, as the subject and object of our experiences, this chapter takes a brief look at the identity of material bodies through time. Although it seems possible to formulate a sufficient condition of this identity in terms of spatio-temporal continuity, the problem is to formulate this condition in a non-circular way. There are also problems about the necessity of such a condition, having to do, for instance, with the fact that the disintegration of a thing into parts followed by a creation of the same sort of thing sometimes amounts to a reappearance of numerically the same thing and sometimes does not. It is suggested that this matter is decided partly by pragmatic considerations and, thus, that the distinction between identity and non-identity is less deep than it might be thought to be. The chapter ends by considering what the account given implies the famous ship of Theseus.

Keywords:   identity, material bodies, ship of Theseus, spatio-temporal continuity

ON our pre-reflective conception, naive somatism, we take ourselves to be identical to our bodies because we assume that they satisfy both the ownership and the phenomenal conditions for being the subjects of our experiences. But in fact neither our bodies nor physical things (it has to be physical things) of any other kind play this double role. So naïve somatism cannot be articulated into a philosophically defensible criterion of our persistence. Still, it supplies a rough-and-ready criterion serviceable in everyday circumstances. According to it, our persistence consists in the persistence of a material body. I shall now look into the notion of the persistence of such a body. Although I think that it will emerge that this notion is probably basic and indefinable, this investigation will throw up some findings that makes it a useful prelude to a discussion of the importance of our identity.

A Sufficient Condition of Material Identity

I have here and there in earlier chapters assumed that the diachronic identity of a material thing entails some sort of spatio-temporal continuity. Now, it is a commonplace that for every material thing, m, there has to be some kind or sort, K, to which m essentially belongs, that is, which is such that m must be a K at every time at which it exists. Is it also the case that, if m begins to exist at a time t 1 in a region r 1 and ends its existence at t n in r n, it will have to exist, as a K, at every time between t 1 and t n in some series of regions linking r 1 and r n in space? Presently, we will find that this is not so: the requisite spatio-temporal continuity is less stringent.

In the foregoing, I have employed the commonsensical framework of enduring things that successively exist (in their ‘entirety’) at different times until they cease to exist. Such a framework is presupposed when we speak of a thing (identified as) existing at one time being identical to a thing (identified as) existing at another. This identity is thought to be (p.299) consistent with the thing undergoing a lot of changes in the course of time. Some changes, however, rule out diachronic identity. To ask for the necessary and sufficient conditions for m 1 at t 1 being identical to m n at t n is to ask: what changes between t 1 and t n are such that if and only if they occur, m 1 will not be identical to m n? For instance, are these changes precisely the ones that are incompatible with there being, at all times between t 1 and t n, something of the kind to which m 1 and m n essentially belong?

The three-dimensional commonsensical framework has a rival, four-dimensional conception that in place of the notion of a thing operates with the notion of the whole of its existence or ‘career’. Accordingly, any shorter time during this period will be only a ‘stage’ or ‘slice’ of the thing. The thing has ‘temporal parts’, or other stages, making up its existence at other times. In contrast to spatial parts, these temporal parts or stages seem to be instantiations of the same kind as the thing of which they are stages, for example, a stage of a ship or sheep is apparently itself a ship or sheep. If the duration of a ship or sheep had been shortened from twenty years to twenty minutes, the result would still be a ship or sheep. In this four-dimensional framework, the problem of diachronic identity will take this form: what conditions are necessary and sufficient for different stages being stages of the existence of one and the same more lasting thing?

The four-dimensional framework could in this fashion be used to rephrase the issue. But it should be borne in mind that the existence of a thing is not the same as the thing itself. For instance, a (material) thing is composed of matter, but its existence is not; its existence has duration, but the thing itself does not. And a thing cannot be identified with the whole of its actual existence, since the thing could have existed for a shorter, or longer, period and still be the same thing. It follows that we need to make up our minds whether a ‘stage’ of a thing is a shorter bit of its existence or the thing considered as existing only at this time. If, however, we are not guilty of these confusions, there may be no harm in employing the four-dimensional framework in discussing diachronic identity.1

It may seem that we must take the relata of the relation of diachronic identity to be momentary, that is, the times at which things are identified to be moments, times having no duration or extension. For if they have extension, one can distinguish an earlier and a later part of the things existing during them which are related precisely in the manner to be analysed. Within each of these parts, one can in turn separate an earlier and a later part, and so on. It is only if we at last arrive at something momentary, and the analysis is applicable to it, that we have succeeded in giving a general, non-circular analysis of what makes a thing persist or retain its identity through time.

Unhappily, there seem to be serious problems besetting such attempts to understand transtemporal identity or persistence in terms of relations between momentary things.2 But even if, in response to these difficulties, we scrap the notion of a momentary thing and grant that, however far the regress is pursued, the relata will be of some, albeit very short duration, it does not follow that the explication, though non-reductive, will be vacuous. It can be informative to be told that two things existing at different times are (p.300) identical if and only if they stand in a certain relation R to each other, although these relata are persisting things that in turn are divisible into things that stand in R to each other, and so on ad infinitum. (Compare: it can be informative to be told that someone is a human being if and only if both of his/her parents are human beings.)

At most, what we can aspire to do may well be, then, to spell out such a relation R that makes identical two things which themselves persist for some period. As already indicated, it is often suggested that the persistence of a material thing, m, consists in the spatio-temporally continuous existence of something of the kind of which m essentially is, that is, that this relation is R. Granted, since the notion under analysis is very pervasive, it is hard to get rid of suspicions that it crops up in various places in the analysans, reducing it to circularity. For instance, it has been argued both that the requisite place-identifications presuppose the identity of persisting objects and that the notion of an essential kind does.3 But, in line with the concession in the foregoing paragraph, let us waive such worries and merely ask whether a continuity analysis along these lines could give a condition that, albeit non-reductive, is both necessary and sufficient for two persisting material things being the same K.

An obvious, and serious, difficulty with this analysis as a necessary condition—a difficulty to which I shall return later in this chapter—is that a thing may fall to pieces without the thing's identity being definitely obliterated. If so, then for m 1, existing in r 1 at t 1, to be the same K as m n, existing in r n at t n, there need not be a K at every time between t 1 and t n in some series of regions connecting r 1 and r n.

Another difficulty for such a necessary condition concerns the matter of precisely specifying the relevant spatial path. This is due to the fact that, from one moment to another, a thing may lose or acquire large parts while retaining its identity: for instance, a big branch could be chopped off a tree without its ceasing to be the same tree. This would, of course, make it occupy a different region, even if it is immobile.4 Clearly, it is indeterminate how much of a thing could be lost without it ceasing to exist. This would be true even if the issue was a purely quantitative one of the size or mass of the parts at stake, but it is further complicated by the fact that parts are often more or less central to a thing (e.g. the trunk is more central to a tree than the branches).

Some have thought it paradoxical to identify, for example, a tree, T, with the tree that exists after a branch has been chopped off it. Suppose that the branch is cut off T at t. Then, it might be urged, the tree existing after t,T*, must be identical to an undetached part of the tree existing before t, namely, this tree minus the branch, TB. For all their parts are identical. But this undetached part, TB, is not identical to T existing before t which possesses the additional branch, B.

I reply by denying that T* is identical to TB rather than to T.T and T* are things of the same kind, trees, so they can be the same thing of this kind, the same tree. In contrast, TB is not a tree, but a proper part of one, which T* is not. What about the claim that T* must be identical to TB rather than to T, since T* and TB share all proper parts (p.301) (while T and T* do not)? As shown by other cases, like that of the ship of Theseus discussed below, the fact that the parts of x at one time and of y at another are identical does not entail that x and y are identical, even if they are of the same kind, as they are not in the present case. If T* were identical to TB, it could not survive the loss of a further branch which it clearly can. Instead, TB wholly composes T*, whereas it partly composed T.

More fundamental is, however, the problem that a condition to the effect that there be something of a certain material kind without any spatio-temporal interruption seems not sufficient for the transtemporal persistence of a single material thing or body of this kind. Sydney Shoemaker has devised a thought-experiment to this effect:5 he imagines there to be both machines that instantaneously destroy tables and machines that instantaneously create them. Suppose that a ‘table destroyer’ annihilates a given table (along with its constituents) at t, but that a ‘table producer’ creates a qualitatively indistinguishable table on the same spot, r, the very next moment. Then there will continuously be a table in r, but, as Shoemaker—to my mind correctly—maintains, it will not be one and the same table. So, a spatio-temporal continuous existence of something of table-kind does not suffice for diachronic identity of something of this kind.

He goes on to argue that an analysis in terms of continuity “has to be replaced or supplemented by an account in terms of causality” (1984: 241). The gist is that what is missing in the situation envisaged is that the fact that there is a table in r just after t is due, not to there being a table in r at t, but to the operation of an external cause, the table-producing machine. If the table existing just after t had been the same table as the one existing at t, one would be able to say truly that there was a table just after t because there was one at t, and not because of any external cause. Following W. E. Johnson, Shoemaker thinks that what is at work here is a special form of causality, “immanent causality”, distinct from ordinary, “transeunt” causality which relates events (1984: 254).

Now, it seems to me to go against the grain to say, for example, that there being a certain sort of table somewhere was caused by there being a similar table in the same place just before. It seems strange to me to hold that such a static state as there being a table in r at t, could be a cause of anything, let alone the state of there being a table in the same place just after t. Instead, I think the full causal explanation of there being a table in r just after t is that at (or before) t a table was placed or created in r, and just after t no cause has as yet removed or destroyed the table in r. There is an external cause of why there begins to be a table in r at a certain time and of why this state ends at a later time. But there is no positive cause of there being a table in r at any intermediate time. The only causal explanation of this seems to be, negatively, that no external cause has as yet removed or destroyed the table. That is, the only possible causes of there being a table in r are external ones. A thing's persistence cannot, then, be defined in terms of any “immanent” causation internal to it.

These observations, however, suggest another way of making the continuity condition sufficient: the table existing in r at t is identical to the table existing in r just after t if there is a table in r just after t because no external causes have removed or destroyed the (p.302) table existing in r at t (and created or placed a table in r just after t). That is, the continued presence of a table does not require any external cause to sustain it, but only the absence of causes that would prevent it. I think this may capture the notion of persistence exemplified in our sense-experience, in particular, in proprioception or the perception of our own bodies from the inside. For instance, the experience I have of the persistence of my body, while sitting and writing this, is of there being a body in a certain chair from one moment to another, without this state being sustained by any external cause.

It may be objected that it is conceivable that tables causelessly cease to exist and pop into existence. That is true, but this possibility requires that there be some discontinuity, I think. It seems no coherent possibility that, without any cause, a table has ceased to exist and another table has popped into existence at the very same time in the very same place.

A graver difficulty is, however, that there are two crucially different ways of ‘destroying’ tables: one that is compatible with their retaining diachronic identity and one that is not. As already indicated, and as we shall soon see in greater detail, an artefact can be dismantled in such a way that it can be ‘resurrected’ if the parts are properly put together again. This means that were we to turn our sufficient condition into one that is necessary as well, we would have to rule out this type of destruction. We would have to specify that the relevant destruction and creation mean that a numerically distinct thing of the same kind would exist instead. But, of course, that would be blatantly circular.

Still, despite its shortcomings the analysis proposed provides some insight. For it brings out a difference between the continuity of a physical thing or body—that is, an entity that possesses mass, has a tangible shape, and fills a three-dimensional region of space—which is not causally sustained by anything external, and the continuity of other physical entities or phenomena, like purely visual entities, for example, shadows, and auditory ones, which is causally sustained by something external, namely proper things. In the case of the latter, too, spatio-temporal continuity fails to make up a sufficient condition, but here the extra element can probably be understood in terms of the continuity being causally sustained by one and the same external thing. So, as in Chapter 20 I suggested about the identity of the mind, I now suggest that the identity of these phenomena is parasitic upon the identity of proper things which may be basic and indefinable.

To simplify matters, we might as well formulate the sufficient continuity condition at which we have arrived in the following openly circular fashion:

(C) m 1 existing at t 1 in r 1 is identical to m n existing at t n in r n if, at every time between t 1 and t n and in some series of regions joining r 1 and r n, there is something of the same kind as m 1 and m n to which they are both identical.

The Pragmatic Dimension of Material Identity

But, as already indicated, we face the problem that (C) fails as a necessary condition. Imagine that a ship that exists at t 1 is dismantled at t 2 and that the planks and so on are stacked away. Some time later, at t 3, a perfectly similar ship is built out of these. It would (p.303) be quite natural to assert that the original ship which existed at t 1 has been rebuilt, that the ship existing at t 3 is numerically the same ship as the one that existed at t 1. But this shows that (C) is not necessary, since the ship is identical to no ship at t 2.

It will not do to retort that the ship is identical to something at t 2, namely a heap of planks and so on: that this is a ship, though a ship in pieces. For suppose that a ship had never again been constructed out of these. Then we would not say that the ship which existed at t 1 exists as long as the heap does; clearly, we would hold that it went out of existence at t 2 when it was dismantled. So we should admit that this example demonstrates that (C) will not do as a necessary condition.

It might be suggested that a simple revision of (C) will take care of the difficulty: suppose we add to it the disjunct (D) ‘or a greater number of the parts constituting m 1 exist at every time between t 1 and t n in regions linking r 1 and r n, and no later than t n they have been joined to constitute m n that is of the same kind as m 1’. I do not think, however, that this revision will meet all difficulties, for we do not believe that a thing retains its identity whenever its constituents, after being separated, later come together to compose something similar in kind.

Suppose, for instance, that the elementary particles composing the ship were dispersed, but that some time later they resume their original positions and constitute a similar ship. This sequence is, we further imagine, irregular: sometimes something of the same kind as the decomposed thing is created, but quite often this does not happen. When it happens, it happens long after the dispersal. Under these circumstances, I think we would be hesitant to identify any ship later created with the one which earlier disintegrated. We would be uncertain whether the original ship has reappeared or whether it been replaced by a new ship resembling it.6

Why this uncertainty? I do not believe that the reason for it is that the disintegration is more radical here, that the constituents into which the ship is decomposed are smaller here. Nor do I think that the length of the interval between annihilation and creation is the decisive factor. Instead I think the reason for hesitancy is that the radical disintegration is a process that we do not control and we cannot predict its outcome (whereas removing planks from a ship and putting them back is of course a process under our control). Imagine that the outcome of the radical disintegration was regular and predictable, that the particles dispersed always rejoin to compose a qualitatively identical thing, or that our knowledge of micro-cosmos grew to the extent that we learnt a method of steering the paths of particles so that we could at will cause macroscopic objects to come into and to go out of existence. Then I think we would be willing to proclaim the thing that later appears identical to the one earlier decomposed, even if it takes a long time to appear. This is, I think, the explanation of the readiness of some SF-writers and philosophers to speak of things being ‘tele-transported’—a description that presupposes identity—when they are dissolved into their micro-constituents and these are sent in a stream to another place where they are reassembled into their original structure (see e.g. Tye, 2003: 147).

(p.304) If this speculation is correct, it suggests that there is a pragmatic dimension to identity across time. It is not the nature of a particular sort of disintegration itself which determines whether it is destructive of identity; it is our knowledge and control of it. When we learn to control a process of decomposition so that we can recombine the parts it separates from each other, or at least reliably predict that such recombination will occur, we are prepared to speak of the original thing as coming back into existence, in spite of our earlier being inclined to see this sort of decomposition as destroying the original thing. This consideration is noteworthy because it supports a main contention of mine—to which I shall come in Chapter 23—namely that diachronic identity is not a deep relation that carries any greater importance.

There is another consideration showing the disjunct (D) to be inadequate. Imagine that, before building a ship at t 3, we had used its components to build a shed; then I think we would be inclined to take this intermission as ruling out identity between the original ship and the new one. But (D) does not harmonize with this verdict, for the parts of a ship continue to exist even if they make up a shed.

The hitch seems to be that a shed is a structure which rivals that of ship: nothing can be both a ship and a shed. In contrast, something may be both a collection of planks and a ship because the planks constitute the ship. Furthermore, in some conditions there is a tendency to identify the planks as ‘parts of a ship’ even when they do not constitute a ship—but not any longer when they have been used to build a shed. Then the possibility of numerically the same ship reappearing is also ruled out. If we did not take the interference of rival structures to block identity, we might, for instance, hold ourselves to be reappearing if, after millions of years during which the elementary particles composing us have helped to constitute other organisms and inanimate objects, they predictably come together to make up organisms indistinguishable from our organisms as they were at the time at which they were disintegrated. But this is surely absurd.

A necessary and sufficient condition that captures these complications would run along the following lines:

(C*) m 1 existing at t 1 in r 1 is the same K as m n existing at t n in r n if and only if:

(1) m 1 and m n are identical to something K existing at every time between t 1 and t n in regions connecting r 1 and r n or, if (1) does not hold:

(2) most of some set of parts constituting m 1 exist at every time between t 1 and t n in regions linking r 1 and r n, and are parts which (a) predictably and no later than t n are united to constitute m n that is a K just as m 1 is, and which (b) have not between t 1 and t n constituted anything of a kind incompatible with K.

In (C*), (2) is intended to be a subordinate condition which comes into operation only if (1) is not satisfied. For this reason the ship of Theseus poses no difficulty for (C*). According to the story, the planks and other components of this ship, s 1, existing at t 1, are gradually replaced by new ones, so that the ship s 2 existing at t n has none of the original parts. These original parts have, however, been hoarded and a ship similar to the original one, s 3, is built out of them at t n. Now which of s 2 and s 3 is identical to s 1? According to (C*), (p.305) s 2 is. The fact that if s 2 had not existed, we would have identified s 3 with s 1 is accounted for by the conditions (a) and (b) of (2) which come into play when (1) is not met.

This does not violate the intrinsicality of identity (mentioned in Chapter 20), for whether s 1 is identical to s 3 does not depend on s 1's relations to anything existing simultaneously with s 3, at t n, but on what happens in between t 1 and t n. This is shown by the fact that, if just before t n s 2 had been destroyed, s 3 would still not be identical to s 1, despite the fact that it has no contemporary competitor for identity. Therefore, psychologists of identity cannot defend their rejection of the intrinsicality thesis by appealing to its breakdown in cases like that of the ship of Theseus.

I do not want to insist on the details of (C*), since I believe identity across time to be a fluid concept which cannot be captured by any precise necessary condition. But I do want to emphasize that the diachronic identity of a material thing consists in the spatio-temporally continuous existence of material things of some kind. In the first instance, this condition takes the form described in (1): an earlier thing is identical to a later thing if it is identical to something of the same kind existing at every time in between. Whether or not this condition is fulfilled is sometimes indeterminate, since, as we have seen, it is not clear how much of a thing can be lost without it going out of existence.

If this condition is not satisfied, we can fall back on a second form of continuity: continuity as regards the existence of components. Were we to give up this constraint, it seems that the notion of diachronic identity has been loosened to the point of dissolution.7 When this continuity is responsible for the identity of the whole, identity becomes even less determinate since, apart from the indeterminacy as regards the requisite extent of the persisting parts, there may be uncertainty as to whether a disintegration–reconstitution sequence is of the right sort. This fuzziness of the distinction between the identity and distinctness of material things is worth underlining because it supports a main claim of this part of the book, to wit, that our identity in itself is nothing of importance. (It will support this claim even if we accept a psychological account of our identity, since, as I contended in Chapter 19, psychological accounts must assume a matter-based form.)

Of relevance for my argument for this claim in Chapter 23 is also the leading idea behind (C*), that continuous existence of things of some kind is a sine qua non for diachronic identity. Every macroscopic thing is constantly involved in a process of having its microscopic constituents successively replaced. Generally, as the size or number of the parts replaced at one go increases, persistence becomes more and more dubious. There is no definite point at which a replacement is sufficiently extensive to bring the whole thing out of existence. (Nor is there a definite answer as to how short the time-interval between each permissible replacement in a series may be without destroying identity.) But if all of a thing's constituents are replaced at one shot, it certainly goes out of existence. Any macroscopically indistinguishable thing that succeeds it will then be numerically distinct from it. To repeat, if it is affirmed that numerically the same thing (p.306) can make a comeback after everything of it going out of existence, it is hard to see what distinction there can be between such a comeback and it going out of existence to be replaced by a numerically distinct, qualitatively identical thing.

Since this is of some importance for what follows, let us take a somewhat closer look at someone who adopts a contrary position. Andrew Brennan imagines scientific investigations showing that in what is apparently a hill enjoying continuous existence, there occur brief intervals of non-existence (1988: 92 ff.). Still, this gappiness would, Brennan claims, prevent us neither from saying, in his favoured terminology, that the hill survives nor from holding the different hill appearances to be causally connected.

But Brennan's view leaves it unclear how he can uphold the distinction between one and the same thing continuing in existence and its ceasing to exist and to be replaced by another, similar one. I cannot discern a more plausible basis for this distinction than one that makes it turn on the presence or absence of the spatio-temporal continuous existence, if not of the thing itself, then of components of it. It seems reliance on causality will not do the trick, for one also wonders how the hills on opposite sides of the discontinuity are supposed to be causally linked. It does not seem intelligible to suppose that the former hill causes the latter to exist before it goes out of existence, a while before the latter begins to exist.

Secondly, Brennan's view forces one to surrender the intuition that diachronic identity is an intrinsic relation between the relata. For suppose that a hill disappears at t 1, but that a similar hill appears in the same place a fraction of a second later, at t 2. Brennan claims that such a discontinuity does not necessarily rule out identity between the hill existing at t 1 and the one being present at t 2. But it is conceivable that another, perfectly similar hill appears elsewhere at t 2. The hill existing at t 1 can scarcely be identical to both of these hills; so, if one cannot come up with a reason for identifying it with one rather than the other, identity is excluded in this case of duplication. But then, contrary to the intrinsicality thesis, the identity of the hill at t 1 and the one appearing in the same place at t 2 depends on the relation of the first to what happens in other places at t 2, on hills not suddenly coming into existence in them at that time.

Brennan can hardly escape this conclusion by contending that the fact that one hill appears in the same spot as the one that disappears presents a conclusive reason for picking out that hill for identification. For he allows that a thing may ‘hop’ not merely in time but in space as well, and it would indeed seem arbitrary to permit jumps in one of these dimensions, but not in the other.

For such reasons, I insist that the diachronic identity of a thing involves spatio-temporal continuous existence at least in respect of components composing it.


(1) But see David Oderberg who argues that “it is precisely the conflation of a persistent with its life-history which permits the stage-theorist to give the appearance of revealing the existence of a novel ontology” (1993: 127–8).

(2) See Saul Kripke's unpublished, but widely known, lectures on identity over time.

(3) See, for example Oderberg (1993: 6–10 and 50–2 respectively).

(4) For further discussion of this difficulty, see Hirsch (1982: ch. 1).

(5) See ‘Identity, Properties, and Causality’, repr. in Shoemaker (1984).

(6) So, we have no unwavering faith in Hirsch's “compositional criterion” (1982: 64–71).

(7) Thus I am unmoved by Hirsch's speculation (1982: 216 ff.) that numerically the same macroscopic object can go on existing even if there is a simultaneous replacement of all or most constituents on the micro-level.