scientific explanation necessity and contingency

7/27/2019 Scientific Explanation Necessity and Contingency

1/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY

Erik Weber

1. IntroductionIn this paper I shall discuss the relations between contemporarytheories of scientific explanation and general ' ' images of science"(i.e. more or less ar.ticulated views on the essential character i s-tics of scientif ic theories) . More specifically, the resul ts ofs tudies in the domain of explanation win form my basis forcriticizing a generally accepted image of science, viz. the ideathat science is a system of natural (or social) laws that expressrelations of physical (or social) necessity between subsequent orsimultaneous events. In general , images of science may be iden-tified by means of the modalities that are crucial in it. To clarifythis, i t is useful to give the following definit ions:DA = the occurrence of A i s necessaryD'A = O",A (= the occurrence of A is impossible)OA = NONA (= the occurrence of A is not impossible)O'A = NOA ( the occurrence of A is not necessary)ilA = NONA&",OA (= the occurrence of A is contingent)Science then may be seen as a system of phrases of the form " I fAt then DB". Scientific progress means to move from a situationwhere ilA to a context in which DA. The opposition between themodalities DA and M therefore is crucial in the image of scienceI jus t sketched. Throughout this paper, i t will be necessary torefer to o the r images of science. For sake of clari ty, I shal ldefine them by re fe r r ing to a crucial opposit ion between twomod ali ties, as I have done here.

My main thesis is that the image of science I described is notcompatible with any adequate theory of scientific explanat ion.Arguments for this thesis are given in section 2 of this paper .Section 2.1 includes some general remarks about the s t ruc tureand the components of theories of explanation. In section 2.2, Idescribe the re levant aspects of two such theories: Hempel's


2/19

82 ERIC WEBERDN-IS model and Salmon's causal approach. In section 2.3, themain argument is developed: I shall point out that the theories ofHempel and Salmon lead to images of science in which thecontingency-necessity opposition is not essential. For both theories, I shall propose a characteristic pair of modalities to replacethis opposition. In section 2.4, I will clarify my thesis by makinga complete turn in the argumentation: I'll s tar t from a theory ofexplanation which corresponds to the cont ingency-necessi tyimage, and show tha t such a theory is always inadequate.

Section 2 deals with science in general. But since the subjectof this volume is above all social science and history, section 3will deal with two major sociological research traditions. The aimis to show that our general conclusions are valid here too, andto give these general conclusions more substance.

2. Explanation and modalities: general account2.1 Elements o f a theory o f scientific explanationMy start ing point is the theory of scientific explanation that isdeveloped by Bas Van Fraassen in "The Scientific Image". According to Van Fraassen, an explanation is an answer to aquestion (Q ) of the form "Why Pk?", where Pk states the phenomenon or fact that is to be explained (i.e. the explanandum). Pkis also called the "topic" of the why-question. But why-questionscannot be identified completely by means of their topic alone.The question "Why did John steal a car?" may have differentinterpretations: "Why did John (and not Jack) steal a car?", or"Why did John steal a car (and not a bicycle)?". To cope withthis problem, Van Fraass,en introduces the concept of the "cont rast-class" of a why-quest ion. A contras t -c lass X of a whyquestion Q is a se t of proposit ions PI, P2, Pk, .. , Pn where Pk isthe only member of X that is t rue . In our example, the (false)proposition "Jack stole a car" would be a member of the contrast-class, at least for the f i rs t interpretat ion of the question,while "John stole a Lic..:.Yde" would be a member of the contrastclass for the second interpretation. Taking these considerationsinto account, the canonical form of a why-question is: "Why Pk,and not Pl .. Pn (i.e. the res t of X)."The canonical form of an answer to such a why-quest ion is:"Pk in contrast to the res t of X, because A". Of course there aresome conditions to be met to have an adequate answer. VanFraassen 's conditions are:(1) Pk is t rue.


3/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 83(2) In :X, only Pk is t rue .(3) A is t rue.(4) A bears relat ion R to {Pk,X}.The relation R is a relevance relation, which Van Fraassen doesnot specify because he's convinced tha t such relation is nevercontext- independent . So Van Fraassen th inks tha t there is norelevance relation which is valid in all scientific disciplines.Relations lih:e intentionality and functionality e.g. may be import an t in social sciences, but they are of no use in physicalexplanation. Of course there are many authors which disagreewith Van Fraassen on this topic. They maintain tha t there is ageneral relevance relation, valid for all sciences, e.g. a causalrelation between A and Pk (Salmon 1984, Cartwright 1983), statistical relevance (Salmon 1971, Gardenfors 1980), derivabili ty(Hempel).

The two aspects of Van Fraassen's theory we mentioned (viz.contrast-classes as tools for identifying why-questions, and therelativity of the relevance relation) cover two main issues whichany theory of scientific explanation must deal with. To my view,there is one more such issue, so we may say that a theory ofscientific explanation must have the following three components:( l) An analysis of the context of explanation: When do we ask foran explanation? What are the presupposit ions of explanationseeking questions? Van Fraassen 's concept of a contras t -c lassand everything connected with is a par t of such an analysis .(2) In a second par t of the theory, we have to specify therelevance relation between the explanans and the explanandum.According to Van Fraassen, no general answer can be given. Onthe other hand, several other philosophers have proposed suchrelevance relat ions.(3) When we have dealt with the context of explanation (1) andminimal requirements on explanations (2), we may turn ourattention to the evaluation of competing minimally suff ic ientexplanations. Criter ia to select the bes t explanation are to bedescribed. Because Van Fraassen's cri teria are quite complex, Ishall not mention them. A less complex solution is Hempel's:deduct ive derivabil i ty and high a poster ior i probabil i ty.

I t is quite clear, I suppose, that the second component - thedefinition of the relevance relation - is the one that is importantfor u s here. What we mean when we say tha t A explains B,depends on the way in which the relevance relation is specified.In section 2.2, tw o such specifications will be described, resp. tobe found in Hempel (1965) and Salmon (1984).


4/19

84 ERIC WEBER2.2 Hempel and Salmon on relevance relationsIn Hempel's theory of explanation, there are four criteria ofadequacy for explanations: l(1) An explanation is an argument with correct (deductive or

ind uctive) logical form.(2) At least one of the premises must be a (universal or statistical) law.

(3) The premises must be t rue.(4) An explanation must sat isfy the requirement of maximal

specificity.I f we try to schematize this concept of explanation, we get thefollowing well-known schemes:(H1) Lt ... Ln

CI CnE

(general laws)(part icular explanatory conditions)(fact to be explained)

and the statistical version(H2) LI ...Ln

CI ....Cn========rE

(general laws, a t least one statistical)(part icular explanatory conditions)(fact to be explained)

The essent ial characterist ic of Hempel's analysis is tha t eachexplanation is considered to be an argument. I f we translate thisinto Van Fraassen's terminology, we have a question "Why E?"and the answer "E, because C". Such an answer is correct i f andonly i f we can construct an argument with C as singular premiseand E as conclusion. So for Hempel the relevance relation isidentical to the relation of (deductive or inductive) derivability.

Salmon's criticism of Hempel's definition of the relevancerelation may be summarized as follows:(1) Hempel's definition is too restrictive, because it excludes low

probabili ty explanations.(2) Hempel's definition is too loose, because he doesn't take the a

priori probabili ty of the explanandum into account.Ad 1: Hempel requires that the a posteriori probabili ty of theexplanandum (i.e. the probability of E, given C and L) is veryhigh (in scheme (H2) this means that 1-r


5/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 85struction of the form (H2) is an argument for believing that E istrue. Besides some practical problems (for instance: what is thevalue of E?), Salmon has pointed out a more fundamental problem.By means of several examples, he shows that there are adequateexplanations which the a posteriori probabili ty of the expla-nandum is very low. As a consequence, not every explanation isalso an argument .Ad 2: Salmon's second problem is best i l lustrated by the follow-ing examples 2 :"John Jones was almost certain to recover from his cold within aweek, because he took vitamin C, and almost all colds clear upwithin a week af ter administrat ion of vitamin C.""John Jones avoided becoming pregnant dur ing the past year ,for he has taken his wife's bir th control pills regularly, andevery man who takes bir th control pills avoids pregnancy."In both cases, the explanans is irrelevant: colds almost certainlydisappear within a week without vitamin C, and men never getpregnant . So these explanations are inadequate, though theymeet Hempel's requirements perfectly. The reason for this is thatthe a priori probability of E (i.e. the probability of E when C andL are not known) and i ts a posteriori probability are identical.

Salmon introduces the concept of statistical relevance (SR) tocope with these problems. A is statistically re levant for B incircumstances C if and only i f P(BIA,C) t P(BIC), From his(1971) paper , we can derive the following defini t ion of therelation of explanatory relevance:(Sl) C is explanatory relevant for E if and only if

(i) C is statistically relevant for E (in circumstances A)(ii) C is not "screened off" by a third factor B, for which

P(EIA,B,C) = (P(EIA,B)(iii) C preceeds E in time

The f i rs t condition consti tutes the main difference betweenHempel's and Salmon's theory. This requirement solves the twoproblems we jus t described. I t is therefore the main reason whySalmon's theory is to be preferred. The two other conditions arenot crucial, because they are compatible with Hempel's definition(so they may be added as supplementary requirements) .

In his (1984) book, Salmon makes a "causal turn". The rela-tion of explanatory relevance is identif ied with the causalrelation:(S2) C explains E if and only if C is a cause of E.


6/19

86 ERIC WEBERThough the definition of the relevance relation is completelydifferent, the criteria for adequate explanations included in (S1)are still valid. The three condit ions of (S 1) a re now seen asevidence for the presence of a causal relation. In other words:these conditions are Salmon's answer to the question: "How canwe prove that there is a causal relation between two phenomenaC and E?" As a consequence, the causal turn has no implicationsas to the adequacy of explanations: the concrete criteria toeliminate inadequate explanations are the same.

The results of our discussion of Hempel and Salmon may besummarized as follows:(1) We know how the relation of explanatory relevance is defined

by Hempel.(2) We know Salmon's f i rs t definition (1971).(3) We know that Salmon's first definition is superior to Hempel's

approach.(4) We know that Salmon's second definition is also superior ,

because the criteria for the adequacy of explanations do notchange.

(5) What we do not know yet is the meaning Salmon gives to theword "cause". We know how we can, in Salmon's opinion,detect a causal relation and prove i ts existence, but we donot know his definition of causation.

Do we need this definition? We can establish the superiority ofSalmon's theory without using i t . But if we want to discoverwhich image of science Salmon's theory of explanation supports ,i t is necessary to pay some attention to his definition ofcausation.

In Salmon's view, the causal structure of the world has threefundamental aspects3: causal processes, causal interactions andcommon causes. A causal process is the means by which s t ruc-ture and order are propagated (transmitted) from one space-timeregion to other times and places. Causal interactions a re themeans by which modifications of a structure (as it is propagatedin a causal process) are produced. Obviously, causal propagation(i.e. the essential characteristic of a causal process) representsthe conservative aspect of causation (conservation of structure),while causal interaction represents the innovating aspect (production of new structures). An electromagnetic wave propagatingthrough a vacuum, or a material particle moving without any netexternal forces acting upon it are examples of causal processes:in the absence of external influence, the existing st ructure ispreserved. Various sorts of collision, or, in general, the fact thatan external force acts upon a material particle, are examples ofcausal interactions.


7/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 87Our common causal talk may be analysed in terms of causalprocesses and interact ions. If, for instance, we say t ha t a

window was broken by boys playing baseball, we have a collisionof a bat with a ball (causal interaction), the motion of the ballthrough space (causal process) and the collision of the ball withthe window (interaction). I don't think we need common causes(Salmon's th i rd concept) to complete this picture. In Salmon'sview, common causes, like causal interaction, p l aya vital role inthe product ion of s tructure and order . I will not explain herewhy I'm convinced that we don't need this third concept, sincethis is not important for our discussion. What is more importantis that two of the three aspects of causation, viz. common causesand causal interaction, are defined in terms of production: theseaspects are two specific ways in which a s t ructure may beproduced. The other aspect , causal propagation, is defined as"the ability of transmitting marks", which means that the essential characteristic of causal processes is their capability of markt ransmission. Mark transmission (MT) is defined by Salmon asfollows 4

MT: Let P be a process that, in the absence of interactionswith other processes, would remain uniform with ref:pect toa character is t ic Q, which i t would ma!'lifest consistentlyover an interval that includes both of the space-time pointsA and B (A t B). Then, a J!"lark (consisting of a modificationof Q into Q'), which has been introduced into process P bymeans of a ~ i n g l e local interaction at point A, is transmittedto point B if P manifests the modification Q' at B and at alls tages of the process between A and B without additionalin tervent ions .

The elements of Salmon's theory of causation I sketched here willbe sufficient, I think, to allow us to answer our main question,viz. what is the image of science that corresponds to Salmon'sconcept of explanation. Together with the analogue question forHempel's theory , this will be the topic of section 2.3.2.3 Explanation and modalities 52.3.1 HempelI t is typical of Hempel's account of explanations that there areno s t ruc tura l differences between explanatory and predict ivearguments . Let 's re turn to scheme (HI):


8/19

88 ERIC WEBERL: ( x ) F ( x ) ~ G ( x )C: F(a)E: G(a)If, as Hempel requires, Land C are true, (HI) can be used as aprognostic argument, i.e. as an argument which tells us what weshould believe (or should accept or expect). Consequently,Hempel's theory may be in terpreted in terms of deontic mo-dalities, or better: in terms of epistemic modalities ("believe","expect" etc.) which are subordinate to deontic modalities. Forthe fur ther development of this idea, I shall use the followingdeontic modalities:D!A = i t is obligatory to do AO'!A = D!IVA (= it is forbidden to do A)O!A = IVO!IVA (= i t is not forbidden to do A)O'!A = ""O!A (= it is not forbidden to do A)n!A = ,.."O!,.."A&NO!A (it is permissible to do A)In a next series of definitions we make the epistemic modality"expectation" subordinate to these deontic modalities:O*A = i t is obligatory to expect AD'*A = O*",A (= i t is forbidden to expect A)v*A = NO*,.."A (= i t is not forbidden to expect A)O'*A = IVO*A (it is not an obligation to expect A)n*A = IVO* .....A&IVO*A (=it is permissible to expect A)Using this terminology, Hempelian explanation may be describedas follows:

"Starting from a knowledge situation in which O'*E or n*E,we construct an argument which results in a new knowl-edge situation, in which O*E."

We have to conclude that there is no immediate link betweenHempelian explanation and necessi ty or contingency. The onlymodalities that are immediately involved are deontic, and in asubordinate role, epistemic.

The general image of science which may be derived fromHempel's theory of explanation is that of a system of "imperativesof expectation" of the form " I f C, then O*E". It is only a smallstep from the practical concept of an imperative of expectation tothe theoretical concept of a natural law: the natural laws aresupposed to be the grounds for our imperatives of expectation.


9/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 89In other words: "Nature" is the legislative power. In this way, i tis possible to establish as indirect link between Hempel's conceptof explanation and the necessity/contingency image. But, as wealready said, the theory itself does not give us any elements tosupport this interpretation. Replacing O*A by OA is a metaphysical decision, which is to be warranted by ontological principles.

We can now formulate a first, partial conclusion: i f we adoptHempel's theory of explanation (which is actually seriously in -adequate), there is no immediate link between explanation andphysical necessi ty. We need ontological principles to establ ishthis link. In other words: Hempel's view is compatible with thenecessi ty /cont ingency image, but does not favour it .2.3.2 SalmonI f we want to find a modal operator which corresponds toSalmon's concept of causation, our f irst task is to find out what"product ion of a s t ruc ture" (causal interaction or commoncauses) and "causal propagation" have in common. As a matter offact, Salmon uses only two crucial concepts in his definition ofmark transmission (MT): (absence of) causal interaction and the"auto-reproduction of a structure" (or: the "spontaneous reproduction of a structure"). So the problem is: what do "productionof a structure" and "auto-reproduction of a s t ructure" have incommon? The answer is easy if we look a t this problem from atechnical point of view. I f we want to reach a certain goal, we (i)produce some new s t ructures , and (ii) we count on spontaneousreproduct ion of these s t ruc tures (because our aim may be located in other space-time regions). Let's consider the baseballexample again (cfr. section 2.2). Suppose we deliberately want tobreak the window. We produce a collision of the ball with thebat, and count on auto-reproduct ion of a st ructure to cause asecond collision, which is our aim. We may conclude that It/hat"production" and "spontaneous reproduction" have in common istha t they jointly determine whether some goal G is attainable.Therefore, Salmon's concept of explanation may be representedby means of the following modalities:

Att(A)

Av(A)Unav(A)Unatt(A)Nctrl(A)

= A is a t ta inable, i.e. there is a method (acombination of production and spontaneous re -production) tha t resul t s in A.=Att(IVA) (= A is avoidable)= IVAtt(IVA) (= A is unavoidable)= IVAtt(A) (= A is unattainable)= Unatt(A) & Unav(A) (= A can't be controlled)


10/19

90 ERIC WEBERUsing this terminology, Salmon's concept of explanation may beformulated as follows:

"Starting from a knowledge situation in which Nctrl (E) weICXJk for a factor C 80 as to reach a new knowledge situationin which Att (E)."

We may draw a first conclusion, which is parallel to what we saidabout Hempel's theory: there is no immediate link betweenSalmon's concept of explanation and the modal operators neces-sity and contingency. For Hempel's concept, it was possible toestabl ish an indirect link (by means of ontological principles).That this is not true for Salmon's concept, is illustrated by thefollowing implication schemes:0: 0 0 '1 1 ~ ~ '0: 0* 0'*

1 ~ * 1 * ~ ~ ' *Att: Att Av1 N c t r l ~ 1

Unav Unat tIn these schemes the arrows indicate valid derivations, a t leastin the f i rs t two. Indeed, the problem is that the schemes for 0and 0* are valid, and that the same scheme for Att is not. SinceA can't be both necessary and impossible, 0 and 0' are exclusive:the t ruth of one of them excludes the t ruth of the other. As aconsequence, 0 implies , and 0' implies ll'. The other implicationsin the scheme are valid by definition. In the same way, 0* and0'* are exclusive, which makes the corresponding implicationsvalid. In the third scheme, on the contrary, the implications "Att- t Unav" and "Av - t Unatt" are not valid because attainability and


11/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 91avoidability do not exclude each other. "Att(A) & Av(A)" is not acontradiction, and may be indicated as "Ctr1(A)". So moving fromAU to 0 is not j us t a matter of ontological decision: thesemodalities musl nol be in terchanged, because they have completely different character i s t ics . As a consequence, Salmon'sconcept of explanation is incompatible with the necessity / cont ingency image of science.

Our conclusions may be summarized as follows:(1) Even i f we adopt Hempel's theory of explanation (though it is

inadequaLe) there is no immediate link between explanationand physical necessity. An indirect link is possible: Hempel'stheory is neu ra1.

(2) I f we (rightly) s tar t from Salmon's concept of explanation, weare confronted with .a more fundamental problem: Att (A) andOA are not interchangeable in the same way as O*A and DA.The reason for this is tha t Att (A) is in some way "para consistent". As a consequence, Salmon's concept of explanation is incompatible with the necessi ty/cont ingency image.

2.4 The modal conception of explanationIn his "Scientific Explanation .. " Salmon dist inguishes three basicconceptions of explanation: the epistemic conception (e.g.Rempel), the causal-ontic conception (e.g. his own theory) andthe modal conception. In theories of this la t ter kind, i t isassumed tha t the relation of explanatory relevance is identicalwith the relat ion of conditional physical necessity . In otherwords: constructing an explanation means to move from a knowledge si tuation in which liE to a context in which OE. An explanat ion shows tha t the explanandum-event is physically necessaryrelative to the explanatory facts. G.H. von Wright gives expression to Lhis idea as follows6 : "\vhaL makes a deductive-nomologicalexplanation 'explain', is, one might say, that it tells us why E hadto be (occur), why E was necessar.Y once the basis (body ofexplanatory facts) is there and the laws are accepted". I t willobvious tha t such conception of explanation perfectly matchesthe image of science 'we are arguing against in th is paper .Unfortunately for the adherents of this view, a modal theory ofexplanation is always inadequate in indeterminist ic contexts .Statistical explanat ion always leaves room for additional quest ioning, because one has to admit tha t E might have failed tooccur, though, on this particular occasion, i t actually did occur.Rempel tr ied to solve an analoguous problem by imposing a highprobability requirement . The problem with this solution is tha tthe dist inction between high and not high is completely arb i -


12/19

92 ERIC WEBERt rary. The analogon of this solution in a modal theory would bethe definition of concepts l ike "almost necessary" and "nearlyimpossible". But this can' t be done in a non-arbi trary way: eachdividing-line between" almost necessary" and mere contingencyis artificial and completely arbitrary. Therefore modal theories ofexplanation are always inadequate.2.5 General implicationsIn this section, I have argued that Salmon's theory of explanation is incompatible with an image of science in which the modaloperators necessity a contingency play a crucial role. Hempel'stheory, on the other hand, is not incompatible with this image,but does not favour i t either. Because Salmon's theory is superior, our general conclusion is that the results of the analysis ofscientific explanations force us to give up the so-called 'realistic'image of science (characterized by the modal operators 0 and 1:1)in favour of a more instrumentalislic image. This instrumentalistic image must be based on the modal operators "Att" and"Nctrl", or, if we are more tolerant with respect to Hempel, alsoon deontic (obligation) and epistemic modalities (belief).

I am not sure whether Hempel would accept my conclusion,and I am quite sure tha t Salmon would not. As I already mentioned, Salmon dist inguishes three conceptions of explanation:epistemic, causal-ontic and modal. Hempel's theory is classifiedas an epistemic approach. This means, I think, that Salmon wouldagree with my interpretation of this theory. But his own theoryis classified as a causal-ontic conception of explanation. No onecan deny that his approach is causal, but I think his approach israther causal-instrumentalistic than causal-ontic: causation is, ifwe adopt Salmon's definition, a practical concept, related to thedomain of production, and not an ontological concept.

3. Explanation and "modalities in social sciences3.1 Functionalism3.1.1 Analysis of functional explanationConsider the following quest ion-answer pair:

Question: Why is there pulsation of the heart in mammals?Answer: The function of heart-beat ing is to make the bloodcirculate in the organism.


13/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 93In functional explanations, the explanandum is a characteristic ofa system or the behaviour of a subsystem. In the example above,the behaviour of the subsystem "heart" is to be explained. Sincethe question is answered by stating the function of the characteristic or behaviour, i t is assumed that the functional relationship has explanatory relevance. A question which arises immedi..,..ately is whether this new relevance relation can be reduced tothe relevance relations we already know (derivability and causat ion). Therefore we will t ry to analyse functional explanationsresp. from Hempel's and Salmon's point of view. Our final aim isto demonstrate that Salmon's approach is to be preferred in thisarea too, which will allow us to expand the conclusions of 2.3 tothe area of functional explanation.

In the tradition of Hempel, functional explanations are analysed as follows:(Fl) Cl: For system S an equilibrium is reached at time tl.

Ll: For all tx: an equilibrium may be reached a t time txonly if condition N is fulfilled.

L2: Whenever a system has characteristic D, the conditionN is fulfilled.

E: At time tt, system S has characterist ic D.The argument in this scheme is not conclusive. To have a validargument, we must ei ther replace L2 by the s t ronger law L /("Whenever .... fulfilled, and vice versa"), or replace the expla-:nandum by a disjunct ion of E with analogue sentences for theother sufficient conditions of N. I f the second solution is chosen,the explanatory value of the argument is almost reduced to zero,and we certainly didn' t give an answer to the initial question.The first solution is theoretically correct , but in most casesempirically worthless, because laws of the form L2' are seldomfound. The well-known concept of 'functional substi tutes ' is theconceptual translation of this empirical datum. Since both solutions are defective, it is impossible to reduce functional explanations to deductive arguments. This means that Hempel fails tomake functional explanation a subspecies of explanation in general, which is a f i rs t drawback of his approach.

If we look a t functional explanations from Salmon's point ofview, the following descript ion can be given:(F2)Question: Why E? (E = characteristic of a system or behav

iour of a subsystem).Answer: The funct ion of E is N.


14/19

94 ERIC WEBERAn answer is adequate i f and only i f (i) there is a C (= adescription of an equilibrium) which is an effect of N, and(ii) E is a cause of N.

Is functional explanation in this sense a subspecies of Salmon'sgeneral concept of explanation? The answer to this questiondepends on how strictly Salmon's "causal i ty requirement" isinterpreted. Functional explanations are no causal explanations i fthis term is supposed to denote explanations in which theexplanans is a cause and the explanandum the effect. Indeed, N(and C) are effects of E, and not causes. On the other hand,functional explanations always use causal relations in the oppo-site direction (the explanans is always on the "effect side"). As aconsequence, i t is impossible to analyse functional explanationswithout reference to causal propagation and causal interaction.Therefore, I think, i t may be considered to be a subspecies ofexplanation in general.

To my view, there are three reasons why Salmon's concept ofexplanation must be preferred for the analysis of functionalexplanations. The f i rs t reason is that , as we jus t showed,Salmon's general concept is adequate for functional explanationstoo, which is not so for Hempel's concept. The other reasons areparallel to those we gave in section 2.2 when comparing thegeneral theories of Hempel and Salmon: Hempel's analysis is toorestrictive because he uses necessary and sufficient conditionrelations while excluding analogue relations with low probabili-ties. Hempel's analysis is also too wide, because i t does notexclude certain i rrelevancies, similar to the examples we pre sented in section 2.2.3.1.2 Functional explanation and modalitiesThe conclusions with respect to functional explanation and mo-dalities can be given now without much fur ther argumentation,since they are obtained by combining the resul ts of section 3.1.1with the conclusions in section 2.3. We must conclude that(1) Hempel's account of functional explanation cannot be t rans-

lated into the 0* terminology, since (Fl) is not a correctargument. This means that we even don't have a basis fromwhich to jump to an interpretation in terms of D.

(2) I f we star t from Salmon's concept of explanation, the conclu-sion is analogue to 2.3: functional explanations must bein terpreted in terms of the modalities Att(A) and Av(A).


15/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 953.2 Ma terialism

3.2.1 Analysis of materialist ic explanationsFor pract ica l reasons, I shall confine myself to the followingHarxian pr inciples :(Ml) "The production relations are determined by the means of

prod uction".(M2) "The political-ideological ' supers t ruc ture ' is determined by

the socia-economic 'basis ' (i.e. by the product ionrelat ions) . "

To my view, both explanatory pr inciples are compatible withSalmon's concept of explanation. But I think they are to be usedin different kinds of explanation. In my opinion, the f i rs t pr in-ciple must be used in simple causal explanations, where a causeis used to explain i ts effect. The interpretat ion of the principleis tha t there is causal interaction between the means of production and the production relations. The second principle must beused in functional explanations. I ts interpretation is not that thesupers t ruc ture is a mere effect of processes in the basis, bu tthat the supers t ruc ture must be functional with respect to theequilibrium of the socio-economic basis. I don ' t want to getinvolved here in the discussion about the empirical adequacy ofhistorical material ism, or about the correct in terpreta t ion ofMarxist theor ies . Consequently, I shal l not t ry to suppor t myinterpretat ion by giving a l ist of quotations, nor shall I discussthe empirical adequacy of, these theories. What I do want to showhere i s tha t each Marxist principle (not only the two we'redealing with) may be seen as describing ei ther a simple causalprocess in Salmon's sense, or a functional relationship. If someone would maintain tha t (M2) is also to be seen as expressing asimple causal relation, this would not be in contradiction with mymain thesis . Only i f someone would say that there is somethingspecial about materialistic explanations, something that can' t bereduced to causal processes and/or functional relat ions, wewould disagree . The s t ra tegy I am following here is probab lyquite obvious: i f the re is nothing in material is t ic explanat ionwhich we can' t reduce to functional and causal explanations, o urconclusions about explanations and modalities are valid for mater ial is t ic explanations too. But before reformulat ing theseconclusions, I shall elaborate my interpretat ion of the principles(M1) and (MZ) in orde r to make i t more plausible.

In my in terpre ta t ion, (Ml) means that there is causal in te r -action between the means of product ion and the product ionrelations. I suppose that i t is quite clear what is meant here by


16/19

96 ERIC WEBER"means of production". On the other hand, the term "productionrelations" needs some clarification. In his unachieved introduction of "Zur Kritik der Politischen Okonomie"7, Marx analyses therelations between three socia-economic phenomena: (i) the dist r ibution of production instruments, (ii) the s t ruc ture of theproduction itself, and (iii) the distribution of the surplus value(the resul t of the production). According to Marx, the distr ibution of production inst ruments determines the organisat ion ofthe production, which in its turn determines the distribution ofthe surplus value. In a 19th century industrial ised society, thismeans tha t the opposition (at the level of distr ibution of pro-duction instruments) between capitalists (owners) and proletarians (not-owners) determines a second opposition (at the level ofthe organisation of the production) between managers and factory workers, which in i ts t u rn determines the way in whichthese people share in the distribut ion of the surplus value(resp. interest and wages). In Marx' view, these levels are threeways in which we may approach the same phenomenon, viz. theproduction relations. Because of the complete parallelism betweenthe three levels, and because of the dominance of the first level,the terms "production relation" and "distribution of productioninstruments" are interchangeable. This will facilitate the discussion of both (Ml) and (M2).

Let's now re turn to (Ml). In my interpretation, there is, a t acertain moment, an interaction between the means of productionand the production relations: there is a quantitative or quali tative change ("modification" in Salmon's terminology) of tl)e distribution of the production instruments, as a consequence of thetechnical evolution of the means of production. This modificationis propagated till there is a new causal interaction. I think thisis what Marx had in mind, because, in his view, the feudal socials t ructure (i.e. the antagonism between landlords and serfs) isreplaced by the capitalistic s t ructure (the antagonism betweencapitalists and proletarians) as a consequence of industr ialisation, i.e. as a consequence of the use of machines instead oftools in production processes. While certain other characteristicsare preserved (e.g. the antagonism in the social structure) theintroduction of machines causes a replacement of landlords bycapitalists as the main owners of the means of production. As tothe second principle (M2), the functional relations may be determined as follows. The system we are analysing is the socioeconomic basis of a society. As we have seen, this may bereduced to the analysis of one level, viz. the distribut ion ofproduction inst ruments between the members of this society.Such system can ' t function normally unless the way in which


17/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 97the production instruments are distributed is accepted by eachindividual member of the society (i.e. acceptB.nce as a necessarycondition for an equilibrium). Acceptance may be brought aboutby means of external pressure or by means of motivation. External pressure is the function of the political par t of the super-s t ructure . Motivation is the task of the ideological par t of thesupers t ruc ture .3.2.2 materialism and modalitiesBecause historical-materialistic explanations are ei ther simplecausal explanat ions or funct ional explanations, they must beanalysed in terms of the modalities Att(A) and Av(A). This meansthat Marx' theory, like the funct ional ist model, is incompatiblewith the necessi ty/cont ingency image of science.3.3 General consequences for the social sciencesIn the same way as I have done a t the end of section 2, I wouldlike to pay some attention to the general consequences of what Ihave argued for here. I t is clear that most of what I said at theend of section 2, is valid here too: functionalism and materialismare to be approached from a more instrumentalistic point of view.To conclude this paper , I will t ry to indicate how this ins t ru-mentalistic approach must be understood. Traditionally, Marxismis in terpreted as a deterministic theory, describing a necessaryevolution of social phenomena. In the same way, functionalism isoften unders tood as a theory that proves the necessi ty ofcertain social phenomena. From an instrumental ist ic point ofview, these theories become sets of techniques to br ing aboutmodifications in the social s t ructure . The technological implications of a theory are crucial. I f we compare the social scienceswith the natural s c i e n c ~ s , we have to admit that the transformation of pure science into technological application is very poorlydeveloped in the social sciences. I f social scient ists want toreach the same level of application, an instrumentalistic tu rn is,I think, a f i rs t s tep in the r igh t direct ion.

Research Assis tantNational Fund for Scientific Research (Belgium)

Rijksuniversi tei t Gent


18/19

98 ERIC WEBERNOTES1. See Hempel (1962) and (1965). The formulation of the criteria

and the schemes are taken from Salmon (1984) pp. 29-30.2. Salmon (1971). pp. 33-34.3. Salmon (1984) pp. 179.4. Salmon (1984 p. 148.5. The modal concepts Att(A), D!A, and D*A, that will be defined

in this section, are taken from Paul Lorenzen's article in vol.35 in this journal (Lorenzen, 1985).

6. Von Wright (1971) p. 13.7 . This introduction was published, together with some other

manuscripts , as "Grundrisse der Kritik der Politischenbkonomie" (first edition 1939).

REFERENCESHempel, C.G. (1962), "Deductive-nomological versus statistical

explanation", in Feigl. H. & Maxwell G. (eds.) Minnesota Studiesin the Philosophy of Science Vol. III, Minneapolis, Universityof Minnesota Press.

Hempel, C.G. (1965), Aspects o f scientific explanation and otheressays in the Philosophy o f Science, New York, Free Press.

Kitcher, Ph. & Salmon W. (1987), "Van Fraassen on explanation",in The Journal o f Philosophy, 84 pp. 315-330.

Kuipers, T. (1986), "Explanation by specification", in Logique etAnalyse 116, pp. 509-521.

Lorenzen, P. (1985), "Praktische und Theoretische Modalitiiten",in Philosophica, 35, pp. 5-20.

Mackie, J. (1974), The cement of the universe. A study of causation, Oxford, Clarendon Press.

Marx, K. & Engels, F. (1974), Marx-Engels-f4,Terke, Berlin, DietzVerlag.

Merton, R. (1957), Social Theol'Y and Social Structure, Glencoe(Illinois) The Free Press .Parsons, T. (1951), The Social System, Glencoe (Illinois), The Free

Press.Salmon, W. (1971), Statistical explanation & statistical relevance,

University of Pi t tsburgh Press , Pi t tsburgh.Salmon, W. (1984), Scientific Explanation and the Causal Struc

ture of the World, Princeton (New Jersey), Princeton University Press.

Stegmiiller, W. (1983), Probleme und Resultate der Wissenschaftstheorie 1. Erklii.rung, Begriindung, Kausalitiit, 2nd (enlarged)


19/19

SCIENTIFIC EXPLANATION, NECESSITY & CONTINGENCY 99edition, Berlin, Spr inger Verlag.

Van Fraasen, B. (1980), The Scientific Image, Oxford, ClarendonPress .

Von Wright, G.R. (1971), Explanation and Understanding, I thaca(New York) Cornell Universi ty Press.

scientific explanation necessity and contingency

Documents