AN ESSENTIALIST PERSPECTIVE ON THEPROBLEM OF INDUCTION

BRIAN ELLISLa Trobe Unzverstty and Unwerszty of Melbourne

ABSTRACT

If one belleves, as Hume dtd, that ali events are loose and sepa-rate, then the problem of induction is probably insoluble Any-thing could happen But if one thinks, as scientific essentialistsdo, that the laws of nature are zmmanent in the world, and de-pend on the essential natures of things, then there are strongconstraints on what could possibly happen Given these con-straints, the problem of induction may be soluble For these con-straints greatly strengthen the case for conceptual and theoretz-cal conservatzsm, and rule out Goodmanes que znferences basedon alternative descriptions of the world This may not, In itself,solve the problem, but it significantly changes zts nature

1. Introduction.

Hume's problem of mduction is one which anses from hmtheory of the nature of reality It is not, however, onewhich can be solved within his metaphysical frameworkFor Hume's world is one of passive ob jects which, thoughthey may in fact always behave regularly, might well be-have irregularly, or in radically different ways Conse-quently, the laws of nature, which purport to descnbe theregular patterns of behaviour actually occurnng in theworld, must be contmgent, and there is nothmg In Hume's

© Principia, 2(1) (1998) pp 103-24 Published by Editora da UFSC, andNEL — Epistemology and Logic Research Group, Federal University ofSanta Catarina (UFSC), Brami

104 Brian Ellis

world to guarantee the continuation of the regulannes weobserve, or ther extensions into areas we have not ob-served

The identaies of things In Hume's world are 'ride-pendent of what they are disposed to do, 1 e of their causalpowers, capaaties and propensaies The disposinons ofthings are supposed to depend on the laws of nature, whichmight be different in different worlds Thus, things whichare constauted very differently might be disposed to be-have in exactly the same way, if the laws of nature weresuffiaently different (cf the Catholic doctnne of transub-stantiation), while things which have precisely the sameconstautions might not, or might not always, be disposedto behave in the same ways in the same circumstances InHume's philosophy, the ways in which things are disposedto behave are supposed to depend, not on their intrinsicnatures or constautions, but on what the laws of naturehappen to be Therefore, for a neo-Humean, there is nosolution to Hume's problem to be found by considenngwhat sorts of things exist in the world

Nor is Hume's problem solvable within a Cartesian,Lockean, Newtonian, Berkeleyan or Kantian theory of re-ality For on all of these theones the world is essentiallypassive, and the ways In which things are disposed to in-teract are contingent upon the laws of nature Things, asthey are in themselves, are supposed to have no genumecausal powers Tf they appear to be causally active or Inter-active, then this is not due to any powers which they mayhave by nature, or which they may have acquired, but tohow they are required to behave by the laws which governthem Metaphysically, things in the world are to bethought of as puppets which are pushed around by theforces of God or nature They are not themselves actors onthe stage

An Essentialist Perspective on the Problem of Induction

Induction is therefore a problem for the broadphilosophical tradition which has rts roots in Seventeenthand Eighteenth mechanism It is not just a problem of em-pincism It is a problem for anyone who believes that thelaws of nature are supenmposed on a world which is essen-nally passive, and that these laws are contingent, and notknowable a priori

To solve the problem of induction, it may be neces-sary to break with this whole way of looking at thingsWhat may be required is a conception of reality which de-rues that nature is essentially passive, or one which acceptsthat events may be necessanly connected For Hume's the-sis that ali events are loose and separate', and conse-quently that there are no necessary connecnons betweenevents, is an ontological daun which, if true, would seemto make the problem of induction insoluble

Of course, a solution to the problem of inductionmust ulnmately be epistemological That is, a must vinda-cate most of our ordinary scientffic inductive practices Buthow we ought to reason about the world might well de-pend on what kind of world we think it is For this will af-fect what we think the epistemic task is If we think theworld is a Humean kind of world, i e a world of discon-nected events, then maybe no inductive strategies can bevindicated If so, then perhaps this is suffiaent reason tobelieve that it is not, after all, a Humean kind of world If,however, we think it is a highly structured, and necessanlyconnected, world of natural kinds of objects, properties andprocesses, as saentffic essentialists do, then our aim mustbe to discover its nature and structure And this is a verydifferent task from that which confronts the Humean

The aim of this paper is to argue that the problemof induction looks very different from the perspective of ascientffic essennalist From this perspective, ali scientffic

105

106 Bnan Ellzs

inference is seen as depending ultimately, not just on ob-served regulartnes, but on what postulates about naturalkinds are justifiable The role of postulates about naturalkinds in scientific reasonmg has been stressed by a numberof wnters in recent years (Butts 1977, Forster 1988, Mac-namara 1991) Most recently, the point has been made byHoward Sankey in his Induction and Natural Kinds'(1997), and by Hilary Kornblith in his Inductive lnferenceand its Natural Ground (1993) Kornblrth and Sankey arguethat the world has a basic natural kinds structure, and thatthis is important, because it gives substance to the idea thatnature is uniform Speafically, it guarantees that certainpropernes are uniquely dustered This is true, and impor-tant, but I thmk that a great deal more has to be saidabout the natural kinds structure of the world, if we are tohave a full appreciation of its role

The structure I envisage is much more comprehen-sive First, it indudes hierarchies of natural kinds of prop-erties and processes, as well as the usual hierarchy of natu-ral kinds of objects Hence, I would extend the range ofphenomena accessible to natural kind reasonmg in newways Secondly, I would postulate that the natural kmds ofpropernes include ali of the physical quannties which arecharacteristic of things, as they are In themselves, and so,presumably, may be amongst the essential propernes ofthese things, qua members of natural kinds Thirdly, I as-sume that the essential propernes of natural kinds includemany that are dispositional, that is, properties which de-termine how, or define the objective probabilmes withwhich, things of these vanous kinds would be disposed toact, and react, in vanous kinds of arcumstances Fourthly,I assume that the natural kinds of processes that we call'causal processes' are just the displays of the intnnsic causalpowers, capacities and propensines of the things mvolved

An Essentialist Perspective on the Problem of Induction

in them, and consequently, that these processes aregrounded in these properties 'The laws of nature which de-senhe the natural kinds of processes which can occur arethus grounded In the intrinsic, dispositional properties ofthings

Some of what I have to say in this paper derivesfrom an earlier paper of mine 'A Vindication of ScientfficInductive Practices' (1965), in which I argued for the virtueof conceptual and theoretical conservatism in inductivereasoning I appealed to such considerations specifically todeal with Goodmanesque problems ansing from the possi-bility of radical reconceptualisations of the world (Good-man, 1955) I argued there that there were very good prag-matic reasons for these forms of epistemic conservatismSpeafically, I argued (1) that theoretical involvement is anecessary condition for the possibility of rational non-demonstrative argument, and (2) that Goodmanesque con-ceptual revisions In a given field could easily destroy thetheoretical involvement of the terms we use to desenhe thethings that exist In this field Therefore, to make such revi-sions unnecessanly is irrational, because it must leave uspowerless to argue inductively in the area of these revi-sions

However, I had no wish disallow conceptual ortheoretical revision altogether I could only insist that it bethorough, and prima facie compatible with any other theo-nes we might have about the kinds of things we think weare dealing with But then, I had to allow that the con-struction of radical alternative theoretical frameworksmight not be ali that difficult, especially if there were noconstraints other than empincal ones on the kinds of theo-nes that could be developed lhe supposed virtues of epis-temic conservatism could thus quickly lose their appealTheoretical involvement may be a necessary condaion for

107

108 Bnan Ellis

the possibility of rational non-demonstrative argument Butperhaps Goodmanesque theones could easily be developedto replace the standard ones, and hence serve as a basis forGoodmanesque inferences

I now think that I can advance the argument fortheorencal and conceptual conservatism a step or two fur-ther For, if the laws of nature are not imposed upon theworld, but anse from the essential natures of things in theworld, then a is metaphysically impossible for things tobehave in any of the bizarre ways envisaged by GoodmanIn his examples For the essential nature of a thing cannotbe dependent on anything that is contingent, such as thedate or place of its existence, or whether or not, or howoften, rt has been observed Consequently, the laws of na-ture, which derive from the dispositional properties whichthings have essentially, cannot contam any references toany such contingenaes Therefore, it makes a great deal ofdifference whether one thinks of the laws of nature as im-positions on a passive world, or as ansing out of rts natureIf they are imposed on a passive world, then anything goesBut if they derive from the essential properties of things,then they cannot be dependent on the specific arcum-stances of their existence

2. Scientific Essentialism

According to saennfic essentialism, the laws of nature arenot connngent, but metaphysically necessary, and they arenot supenmposed on the world, but immanent in rt Todiscover what the laws of nature are, the scientific essen-tialist holds that we must investigate nature empincally tofim:1 out what kinds of things exist, and what propernesthey have Scentific essentialism thus focusses our atten-non upon natural kinds and properties, and how we canknow about such thmgs

An Essenttahst Perspective on the Problem of Induction

lhe main tenets of suentific essentialism that arerelevant here are

1 lhe most fundamental things, properties and processesexisting in the world ali belong to natural kinds That is,there are natural kmds of events or processes (dynamickmds), natural kinds of properties or relations (propertykinds), and natural kmds of objects or substances(substantive kmds)

2 The natural kinds in each category occur in hierarchiesThe most general kmds are category-wide, and permit a\vide range of intrinsic vanation amongst their in-stances The most specific kmds, however, have in-stances which are ali intrinsically identical

3 Substantive things are not passive, but essentially activeand Interactive That is, the properties in virtue ofwhich thmgs belong to substantive natural kmds alwaysmdude some which are dispositional, i e some whichare causal powers, capacities or propensities

4 lhe elementary causal processes in which real disposi-tional properties are displayed ali belong to naturalkinds lhe causal laws of nature describe these kinds ofprocesses, and hence the ways in which things are neces-sanly disposed by their properties to act or interact

5 The laws of nature are therefore not contmgent, butmetaphysically necessary Anything which has a realdispositional property is necessanly disposed to display itin appropnate triggering circumstances

6 The real dispositional properties of things are propertiesin their own nght They are not ontologically dependenton non-dispositional propernes or connngent laws ofnature

109

If scientific essentialism provides us with a true pic-ture of reality, then it gives substance to the idea that na-

110 Brictn Bills

ture is uniform, as Sankey suggests For saentific essential-ism postulates a highly structured world consisting of hier-archies of natural kinds of objects, properties and processes— ali things which, one way or another, are involved astruthmakers for the laws of nature So the task of inquiry,according to an essentialist, is to discover what kinds ofobjects, properties and processes exist, and what their es-sentai properties are And once we know this, we shall alsoknow what the laws of nature are Scientffic essentialismthus provides us with a much more specffic agenda for sci-entific inquiry than any that is provided by the bald daimthat nature is uniform Without a great deal of elaboranon,the thesis that nature is uniform tens us very little As itstands, it would be too indeterminate to be much use, evenif the laws of nature were just regulannes in the behaviourof things, as the Humeans believe What would also beneeded is an account of uniformity, and a justification foraccepting this account rather than any other But such anaccount and justification are notonously difficult to pro-vide

Scientific essentialism implies that to accept the tra-daional Humean view that the dispositional properties ofthings depend on the laws of nature is to get the relation-ship of ontological dependence between dispositional prop-erties and laws of nature the wrong way around For thelaws of nature, according to scientific essentialism, are notregulanties imposed on a passive and obedient nature, butderive from the intnnsic dispositional propernes and con-strtutions of the things that go to make up the world Inother words, the laws of nature are immanent in things,they are not the principies according to which they areregulated

However, to defend scientific essentialism ade-quately, it would be necessary to elaborate a theory of real

An Essentialist Perspective on the Problem of Induction

properties, especally of real dispositional and intrinsicproperties, an ontology of natural kinds, a theory of laws ofnature, and to answer the obvious objections to the daimthat the laws of nature are metaphysically necessary Butthis is a task well beyond the scope of the present paperHere I must ask you just to put aside your objections to theoverall programme of saentific essentialism, and focus onthe question of how one should reason about a non-Humean world which is just as the suentific essentialist be-heves the actual world to be

3. Propernes, Processes and Laws of Nature

Most properties of scientific interest, such as the atomicweight of a chlorme atom, or the charge on an electron, orthe p/h of a standard solution, are quantaative That is,they are ultimately specific, or infame, species of quantities,such as mass, charge or p/h These quantmes are naturalproperty kinds, and quantaative properties, like thoselisted, are amongst the quantaatively specific speaes ofthese kinds 'The most speafic speaes of a given quantityare, of course, to be distinguished from as instances Theinfimic speaes of a quantay are as precise values But twoor more different things could have the same quantay tothe same precise degree So a precise value of a given quan-tay is not an instance of that quantity, unless the quantityaself is a second order universal, as Bigelow and Pargetter(1988) maintain

But Bigelow and Pargetter are wrong For even themost precise value of any quantay may have many In-stances, and if the property of havmg this quantay to thisprecise degree is a universal, as a surely is, then these In-stances must ali be identical Yet the particular objectswhich have this quantay to this degree clearly need not be

111

112 Brtan Ellts

identical A positron and a proton both have una positivecharge, but no proton is identical with any positronTherefore, the instances of a quantitative property are notthe objects or substances which have it, but the tropes ofthat quantitative property in those objects It is importanttherefore to distinguish between the instances of a quantityand its infimic (1 e most specific) speaes The instances ofany quantity are its tropes Its infimic species are its precisevalues

Ontologically, genenc properties, including quanti-ties, are more fundamental than any of their speaes Nospecific value of any quantity, for example, is necessary forthe existence of the quantity rtself Thus, mass might existin a world in which nothmg had unit mass But the reverseis certainly not the case Unit mass could not possibly existin any world in which the generic quantity, mass, did notexist It is to be expected, therefore, that quantities shouldplay a more important role in physical theory than any oftheir speafic values Even a cursory glance at physical the-ory would seem to conf.= that this is so Yet many theo-rists (e g Armstrong 1997) have made the contrary as-sumption, viz that the generic propernes depend ontologi-cally on their infimic speaes But this is an unwarrantedassumption which confuses the ontological order with theepistemic one It is true that we never observe genenc prop-erties directly We observe only the tropes of these proper-nes which are necessanly quite specific in their values Butontologically, it is the generic properties which are funda-mental

The most general laws of nature, which apply towhole ontological categones, are undoubtedly the mostfundamental The conservation laws, for example, encom-pass ali events and process within that scope For whatthey say is that every event or processes which can occur in

113An Essentialtst Perspective on the Problem of Inductton

worlds hke ours is intrinsically conservative of X, where Xis the conserved quantrty More speafic laws, such as thelaws of electromagnensm, are narrower In scope than theconservation laws, and apply only to a more limited rangeof phenomena For these laws teul us nothing whateverabout phenomena which are not electromagnetic ia charac-ter

The laws that we call 'causal laws' are a species oflaws of nature concerned with relations between orderedpairs of natural kmds of events lhe events of the first (orcausal) kind are the triggering events of a causal power,while those of the second (or effectual) kind are the possibledisplays of this causal power More formally, these orderedpairs of kinds of events are such that the occurrence of anymember of the first kind (a cause) will necessitate, or bnngabout, an occurrence of some member of the second kind(rts effect) That is, for every event of the causal kind,which is assoaated with a given causal power, a is neces-sary that there should be an event of the effectual kmdwhich will be as display Any statement describing such anecessary connecnon between natural kinds of events is aqualltative causal law

If the events of the two kinds can be descnbedquantitatively (as they usually can be), then to each quan-trtatively speafic species of event of the first, or causal,kind, there urdi be a probability function ranging overquantaatively speafic species of events of second, or effec-tual, kmd lhe statement of this probabilay function is aquantitative causal law Where we have such a law, a par-ticular event of the causal kmd (which will necessanly bequanntatively spectfic) will, necessanly, make a probable tosome precise degree that an instance of a quanntativelyspeafic species of event of the effectual kind will occur

114 Brian Ellis

4. The Essenttalist's World

Essentialism implies that some events make other events ofvarious kinds more or less probable They rarely necessitateother events of quantitatively specific kinds Consequently,the laws of nature are mostly not deterministic, but prob-abilistic However, to simplify the picture, in order to get adearer view of the essentialist's world, let us suppose thatthe causal laws are ali deterministic That is, let us assumethat events whtch belong to any given causal kind alwaysnecessitate quantrtatively specific kinds of events of thecorresponding effectual kind For example, let us assumethat a triggering event, vis a vis any given causal power, willalways necessitate a specific kmd of display of this power

Suppose, for example, that P is a natural disposi-tional property which would be triggered in arcumstancesof the kmd C to produce an effect of the kmd E Then theprocesses of this kind will themselves constrtute a naturalkind, the essence of which is that rt is a display of P Re-fraction, for example, is a natural kmd of process, the es-sence of which is that a is a display of refractivity Onecould, perhaps, imitate refraction by some elaborate systemof diffraction gratings and mirrors But if it is not a displayof the refractivity of some material, then it is not a case ofrefractton, however like such a display rt may be Moreo-ver, nothing could be a display of refractivity if it did nothave the characteristics which are essential to such a dis-play, or were n.ot produced by the refraction of lightthrough a refracting medium Therefore,

Li For ali x, necessarily, if x has P, and x is in circum-stances of the kind C, then x will display an effect ofthe kind E

An Essentialzst Perspective on the Problem of Induction

Note that the necessity operator in this formula is withinthe scope of the universal quantzfier, i e in de re positionMoreover, if a is an individual which has P necessanly, e gbecause P is an essennal property of a, then we may detachand deduce that

Necessanly, if a is in ercumstances C, then a will dis-play an effect of the kind E

And this, it should be noted, is a necessary connection be-tween events of precisely the kind that Hume rejected

The kind of necessity involved In a causal law suchas this is therefore metaphysical necessity This kmd of ne-cessity has the strength of logical necessity, but is groundedin nature, rather than in language or thought The samereasoning applies generally to ali causal laws For ali suchlaws describe natural kmds of processes which are the dis-plays of causal powers So ali causal laws will be necessarytruths, i e they will be true In ali possible worlds Butmetaphysical necessities, such as these, are unlike otherstrong necessines, such as those of language (analytictruths) and thought (logical truths), because they are dis-coverable only a posteriori They are not a priori necessarytruths, but a posteriori necessary truths which can be foundonly by the emp.rical methods of science

The essentialist's world is therefore not one inwhich ali events are loose and separate On the contrary, ais a world dommated by causal powers, In which eventstriggering these powers necessitate other events which aretheir displays If it were a deterministic world, as I haveimagmed, then any two things which are intrinsically thesame must, of metaphysical necessity, be disposed to be-have in the same way, and, indeed, must behave in thesame way in the same arcumstances Even in an indeter-

115

116 Brian Elhs

ministic world somethmg akin to this is true Two thingswhich are intrinsically the same must, of metaphysical ne-cessity, be intrinsically clisposed to behave in the same kindof way (with a real probability distribution ranging over thepossible expressions of their common intrinsic properties)

lhe essentialist's world is therefore a bound andconnected world If what we take to be the same naturalkind of thing recurs, and we do something of the same sortto II, then we should expect it to respond as any member ofthat kind must respond qua member of that kmd Speafi-cally, it should display the essennal dispositional propernesof things of that kind for which the action we took is atngger If rt does do so, then there is nothing to expiam,except, perhaps, how the process works If it does not doso, then the question anses why should this thmg be dif-ferent from other things of its kind ? There are many possi-bilmes either (a) the thmg does not belong to the naturalkind to which II appears to belong (It might, for example,be a different speaes of the same genenc kind ), or (b) whatwe did to rt was not an effective trigger (i e did not belongto the appropnate natural kmd of triggering events), or (c)we are mistaken about what the essential propernes of thekind were, or (d) the expected effect did occur, but wasmasked by other events, or

So, for an essentialist, the problem of induction hasa rather different flavour It is not a question of justifyingthe mference from 'ali observed As are Bs' to 'ali As are Bs'This inference would be justified automancally if we hadgood reason to believe that the As we had observed be-longed to a natural kmd, and that the property of being aB was an essential property of the As In that case, theproblem would be, rather, to expiam the failure of such aninference Where we have a case of inductive failure, thereare many possible explananons which must be sorted

An Essentialist Perspective on the Problem of Induction

through, and the scientific task is to do this, eliminate al-ternatives, and determine which of the remairung alterna-tives provides the best explanation We might decide thatthe dass designated by 'A' is not a natural kind class, orthat there are no essential properties of As in virtue ofwhich they are Bs, so that if an A is a B, then this is justan accident, or that the apparent exceptions are either notreally As, or they really are Bs (i e monster -barring andmonster-adjustment), or any of a number of other thingsBut whatever we decide, we will have learned somethingfrom our experience, and our conceptualisation of theworld will have been improved

5. Natural Kinds, Cluster Kinds and Social Con-structs

Natural kinds are distinguished from other sorts of thingsIn several ways Specifically, they are objective kinds Thatis, they are kinds which exist in nature independently ofour classificatory systems Their existence does not dependon how thmk about the world, the language we use, whatdistmctions we make, or what classes of things we recog-mse Natural kmds are nevertheless clearly distmct fromeach other, i e they do not merge one mto another, sothat an arbarary distinction would have to be made toseparate them lhe distmctions between natural kmds existobjectively They are there in nature for us to discover,whether or not we recognise them And they cannot bedefmed int° or out of emstence Natural kmds are not onlydistmct from each other, but are distmguished from eachother by an mdefunte number of mtrmsic properties orstructures which are severally necessary and jomtly suffi-cient for membership These properties and structures to-gether constitute what is called the real essence of the kind,

117

118 Brian Ellis

and things which belong to the kind are said to have thesepropernes essennally

Objects and processes of kmds which sansfy theserequirements are the usual objects of study in the physicalsaences, and also in parts of biology, microbiology andbiochemistry There are, for example, thousands of chemi-cal compounds which satisfy them, and tens of thousandsof chemical processes However, the normal objects ofstudy in the social and human saences are not naturalkmds, and the biological distinctions between speaes arenot natural kind distinctions, (in the stnct sense in whichthis term is being used here) Therefore, to the extent thatnatural kind reasonmg differs from reasoning about otherkmds of thmgs, the reasoning involved In theory construc-non in the physical saences will be different from the rea-soning required for theory construction elsewhere It is aswell, therefore, to be cautious about drawmg inferencesfrom the logic of the physical saences to apply to the socialsciences, and conversely The social and physical scienceshave very different kinds of subject matters, and we shouldnot expect their logics to be the same

Biological speaes are what I cal! 'cluster kinds' Withrare exceptions, (e g clones and identical twms), eachmember of a biological species is different intrinsically fromevery other member It has a different genetic make-upTherefore, unless we are prepared to count each member ofa biological speaes (which has no identical twm or done)as sui genens, the species classification is not a strict naturalkmds classificanon For members of the same biologicalspeaes are not constitutionally identical Moreover, thedistinctions between species are not always objective, evenif we confine ourselves to considenng only living animaisand plants If we also wish our dassification system to ap-ply to histoncal animais, as we surely do, then there will be

An Essenticthst Perspective on the Problem of Induction

no objective distinctions between adjacent speaes Never-theless, biological speces are suffiaently like natural kindsfor many of the normal expectations of natural kinds rea-soning to be fulfilled The natural instincts, mating habits,growth patterns, digestive systems, bone and muscle struc-tures, etc of any one group of animais of a given speaesare to be expected to be the same as those of any othergroup of animais of the same species And, if any such in-ference falis, then, barring extraordmary explanations, thismay be suffiaent ground for recognising the existence ofsubspecies, or vaneties, of the speaes in question

In the social saences, the kinds we have to dealwith are mostly soaally constructed, and very few of theobjects of investigation sansfy any of the critena for naturalkmd-hood Therefore, we must proceed in the social sa-ences with far weaker assumptions of umformity than weshould in the physical or biological saences Of course,human beings, as members of a biological speaes, must beexpected to develop similarly, to have more of less similarcapaaties, to respond to situations in similar ways, to bemonvated by the same sorts of considerations, to act simi-larly, and so on That is, we have a more or less commonhuman nature Consequently, there are normality projec-tons we can make, and explanations we can offer, of hu-man responses to social events which have fairly generalvalidity Social events are therefore not quite as loose andseparate as Hume imagmed all events to be Nevertheless,the social institutions we construct as humean beings arenot natural kinds, and so we must be very careful in gener-alising about them 'There are no social forces which have aumform mode of action across socienes, and therefore thereare no natural laws of soaety to be found

119

120 Brtan Ellis

6. The Case for Epistemic Conservatism in Induc-tive Reasorung

In 'A Vindication of Scientific Inducnve Practices', I arguedthat `theoretical involvement [of the subject matter of ourinferences] is a necessary condi-non for the possibilay ofrational non-demonstranve argument' (Ellis 1965, p 296)That being the case, I argued, a is necessary eaher (a) topredict that future observations will conform to acceptedtheones about the nature of the subject matter, or (b) torecast our theones about the subject matter in ways whichare no less acceptable, given past observanons, and to pre-dict that future observations will conform to these newtheones No other strategy will guarantee the preservationof the kind of theoretical involvement necessary for ra-tional non-demonstranve argument

Wahout theorencal involvement, I argued, everyprediction about the future, or unobserved past, can be jus-nfied with reference to any plausible inductive rule For dif.-ferent ways of conceptuahsing what is known to have hap-pened lead to substannally different predictions about whatwill happen, or what must have happened Therefore, ifthere is no preferred way of conceptualising what has hap-pened, then there is no preferred way of projecting whatwill happen Anything goes

Therefore, I argued, it is ai-anona' to project any se-quence of events into the future, or unobserved past, insuch a way that, if the projecnon were actually to be con-firmed, then our theoretical understanding of the subjectmatter of our inference would be destroyed This is what iswrong with Goodman's `grue' and %leen' pro jecnons, forexample If green things hke emeralds, and blue things likesapphires, really did turn out to be grue and bleen respec-tively, and -dm were confamed, not only by observation,

121An Essentialist Perspective on the Problem of Induction

but photographically and spectrographically, then ourtheoretical understanding of matter, and of hght and col-our, would be thrown into turmoil It would leave us In astate of bewilderment, no longer knowing what to expectabout the colours of things in future Therefore, to acceptsuch an inference is irrational To do so is to re ject ourprevious theoretical understanding of the subject-matter ofthe inference, not on the haus of anything that is knownto have happened, but solely on the basis of whilt is projectedwill happen If this were a rational procedure, I argued, thenali theones could be rejected out of hand, and then wereally would be in the a-theoretical position that Humeimagined us to be

I see no reason to disagree with any of this But, atthe time of wraing, I had not seen the connection withnatural kinds Nor would I have quarreled with Hume'sview that anything which is conceivable (or imaginable) ispossible So I had no difficulty with the possibility of thingsbehaving in irregular, or even bizarre, ways For me, possi-May was just epistemic possibility I had no conception ofreal possibility Consequently, what I thought was possiblewould have allowed my imagination free reign My con-straints on irregular or bizarre projecnons of observedregulanties did not derive from a behef that such projec-tions are not real possibilities It was just that we had noviable alternanve to the kind of epistemic conservansm Ithen advocated It was that or the bush

My residual worry was that `alternative, but so farequally sansfactory theones' about the subject-matters ofour inductive inferences, might turn out to be fairly easy toconstruct If so, then nothing would have been gained bymy argument for epistemic conservatism For a worldwhich is, as far as we know, compatible with many differ-ent, but overall coherent, readily constructible, and equally

122 Bnan Ellts

sansfactory, conceptions of realay, would be a world withtoo many chuces It would be a world with too many sign-posts If alternative, Goodmanesque, global theones of re-ality could easily be constructed, then alternanve, Good-manesque, projections of observed regulannes could bemade wahout violating any of the requirements of epis-temic conservansm

7. The Essentialist's Perspective

From my present scientific essentialist perspective, my ear-her worry about the possibility of alternative conceptuahsa-tions of reality now appears to be unfounded When I waswraing then, I assumed, as nearly everyone else did, thatthe laws of nature are contingent, and are supenmposed onan unsuspecting, and passive, world Consequently, 1 hadno argument against the possibility of there being Good-manesque laws of nature, and hence Goodmanesque dispo-sitional propernes I now think that this is ali wrong Thelaws of nature are not contingent, but necessary, and theyare not supenmposed on things, but are immanent inthem The pnnapal scientific task, according to a scientificessentialist, is to discover what natural kinds of thingsthere are, and what their essential propernes and structuresare When we know this, we shall know how things arenecessanly disposed to behave and interact, and thus whatthe laws of nature are

The conceptualisation which has resulted from thisendeavour is the one in current use Therefore, to reject ais to reject one of the principal achievements of science Forthere is no alternative conceptuahsation which is compati-ble with these achievements If there were, then we shouldall have heard about a It is all very well to imagine that wemight ali be brains in vats, and that our scientific knowl-edge is consequently illusory, or that, with diligence, we

An Essentictlist Perspective on the Problem of Induction

could redesenhe the world that is so far known to us in aGoodmanesque way But the first is a sceptical positionwhich has no bearing on the problem of induction, and thesecond is an inference drawn from the alleged arbitrannessof linguistic conventions Neither tens us anything aboutwhat the world is really like Only our science tens us thatHence, the kind of epistemic conservatism that is requiredto justify our scientific inductive practices is not a stancewhich needs any independent justification Epistemic con-servatism of the kind required to justify our scientific in-ductive practices is just the determination to hold on to theachievements of science, including its conceptualisations,until we are forced to make changes to accommodate newinformation If we are not epistemic conservatives In thissense, then we do not believe in the results of science, andthere is nothmg for us to justify

From the standpoint of a scientific essentialist, date-or observation-dependent propernes like grue and bleencannot be charactenstic of kinds For the date of an occur-rence concernmg an ob ject of a given kind, say A, dependsaccidentally on its relanonship to those events which serveas reference points for dating purposes Therefore, therecannot be an essential property of As which is date-dependent, i e there cannot be an essential property of Aswhose mode of operation is a function of the date There-fore, there cannot be a law of nature describing how As areby nature disposed to behave which is date-dependentTherefore, a is impossible, metaphysically impossible, thatanything should be, by nature, grue or bleen Such proper-nes simply cannot exist in any world in which the laws ofnature depend on the essennal propernes and structures ofnatural kmds of things

Sciennfic essentialism thus imposes very strong limi-tations on the kmds of conceptual and theoretical innova-

123

124 Brum Ellts

tions that can be made It is one thing to be constrainedonly by considerations of empincal adequacy, and andefined notion of regulanty It is another to be constrainedby the requirement that things must always be supposed tobehave stnctly according to their intnnsic natures

References

Armstrong, D M 1997 A World of States of Affairs Cam-bridge The University Press

Bigelow, J C & Pargetter, R J 1988 "Quantrues "soplucal Stuches 54 75-92

Butts, R E 1977 "Consilience of Inductions and theProblem of Conceptual Change in Saence " In Colodny,R G (ed ), Logic, Laws and Life Patsburgh PittsburghUniversity Press, pp 71-88

Ellis, B D 1965 "A Vindication of Scientific InductivePractices » American Plulosophical Quarterly 2 296-305

Forster, M R 1988 "Unification, Explanation, and theComposition of Causes in Newtonian Mechanics "Studies m the History and Plulosophy of Science 19 55-101

Goodman, N 1955 Fact, Fiction and Forecast Cambridge,Mass Harvard University Press

Kornblith, H 1993 Inductive Inference and its NaturalGround Cambndge, Mass MIT Press

Macnamara, J 1991 "Understanding Induction " BritishJournal for the Philosophy of Science 42 21-48

Sankey, H 1997 "Induction and Natural Kinds " Principia1(2) 235-54

KeywordsInduction, natural kinds, scientific essentialism, Goodman,epistemic conservatism

Brian EllisLa Trobe University and Melbourne University

B Elhsalatrobe edu au