University of Oxford Department of Politics


Philosophy of the Social Sciences: Lectures

Elizabeth Frazer

Michaelmas 2002



the realist critique of causal explanation



The Problem

Philosophers have had the utmost difficulty in arriving at an operationalisable philosophical analysis of the difference between a genuine ‘cause’ and an ‘accidentally true generalisation’ or simple empirical association.  Yet in social science we need to know whether one factor really is the cause of another. 



Key Terms



In contemporary logic texts often represented by:


A logical relationship between statements or propositions.  An argument entails its conclusion when the conclusion follows deductively/necessarily/logically from the premisses.  Sometimes expressed as ‘therefore’.   The (syllogistic) argument:

P1: All X are Y

P2: All Y are W

P3: No Y are Z

C: All X are W and No X is Z

is one in which the conclusion is entailed by the premisses



In contemporary logic texts often represented by:


A logical relationship between statements or propositions.  An argument implies its conclusion when the conclusion can be said to ‘follow’ by inductive inference from the premisses. 

P1: All the swans so far observed are/have been white 

P2: It is 1603 and the Queen is dead

C: The next swan we see will be white


P: Marriages like Cindy’s very rarely last

C: Cindy’s marriage will not last



There is a form of necessity that governs relations between statements (a logical operator); usually represented by ‘ٱ’ (called ‘box’):


ٱX  (say ‘necessarily X’) is true in any world upon interpretation i iff [say ‘if and only if’]  for every possible world in the set of all possible worlds, X is true in every possible world upon interpretation i. 



A logical relationship between statements or propositions, usually represented by  ◊ (called  ‘diamond’)

◊X (say ‘possibly X’)

Note that this is equivalent to:

¬ ٱ¬X




Hume was a key figure in ‘associationism’ – an eighteenth century analysis also associated with the work of John Locke  (1632-1704)  and  James Mill (1773-1836).  The starting point is the striking way our ideas (mental images etc) are associated with but independent of sense experiences,  and the way our ideas seem to be systematically associated with each other.  Associative laws for long dominated the philosophical analysis of psychology, and also generated the beginnings of experimental psychology.  It became, in Lakatos’s words, though, a ‘degenerating research programme’.


From our point of view the significance of association is that the relationship between the two entities associated is nothing more than co-occurrence or correlation – the concept of association carries no connotation or denotation of there being any mechanical connection or any causal connection, between the two entities.



1.             Hume  analyses our idea of cause as having three distinct elements:

C causes (C->) E if E follows C;  E and C are contiguous or connected by a chain of contiguous events, and  there is some necessity in the relationship between the two.

Hume offers two analyses of cause, implying that they are alternative formulations of the same relation.

‘we may define a cause to be an object, followed by another , and where all the objects similar to the first are followed by objects similar to the second.  Or in other words where, if the first object had not been, the second never had existed.’ (Enquiry Concerning Human Understanding p76)

Note that the two propositions are not equivalent. 

P1. C ^ E

P2. All C (C ^ E)

P3. (¬C) ^ (¬E);    or     (¬C) → (¬E)


2.             The issue for Hume is whether our idea of cause has any rational justification.  His argument is that

C, ¬E

 – not just because, say, a lower branch supports the cradle in the stead of the one that breaks; but because there is no logical contradiction in the ‘laws of gravity’ reversing themselves. (cf EARLIER discussion of Hume on induction) 


Hume’s argument then

·         denies the rationality of our belief in any necessary connection between those events in the world we consider to be causally connected. 

·         but, that our belief is not rationally based is not a denial that we do have such a belief – its foundation is psychological (as is the foundation of our use of inductive inference, belief in the continuity of past into future etc.)


Hume anticipates a good deal of Mill’s analysis of  methods for establishing that relations between objects, events or phenomena are causal.  In the Treatise Hume offers ‘Rules by which to judge of cause and effect’ which anticipate Mill’s methods of agreement, difference, and concomitant variation.  But these ‘rules’ pertain to our minds – they address the question ‘when do we believe that two events are causally connected?’


3.             Hence the shift to logic. (Psychology is not an adequate foundation for scientific knowledge)


4.             However, the philosophical analysis of cause runs into the difficulty of the ‘truth value’ of ‘counterfactuals’. 


we have a number of

·         background conditions:  Oxygen present, atmosphere dry, ...

and an infinite number of

·         counterfactual conditions: there was no gale blowing,  no human being present blowing against the spark, no fireman directing a firehose at the spark, ...


5.             As if that’s not enough the analysis of cause then runs into a number of difficulties about

[These issues will be  taken up in the next lecture.] 


6.             More recently a number of philosophers have attempted to analyse cause in a way that is realist about causal connections.  One strategy has been to avoid analysis in terms of ‘events’ (which are discrete) and shift to analysis of processes which are non-discrete.  This may be thought to elide the gap between C and E into which  Hume’s scepticism gets its hook. 


1) C: the brick hitting the glass, E: the window breaking;

2) C: the matchead striking the rough surface, E: the spark occurring;

3) C: the branch breaking, E: the cradle falling

4) C: the sugar dissolving in the water; E: the syrup forming

Perhaps the world is more like four than it is like one to three.

 Or, can we reconceptualise 1-3 as like 4:

4) the sugar dissolving in the water is the syrup forming;

5) the water temperature falling to zero and below is the ice forming. ??




In the context of the philosophy of science ‘realism’ signals a number of analyses and approaches:


Who has said what about this?



See above.  That’s enough Hume – Ed.



Constructed logical rules for determining when, in science, we may justifiably conclude that two events  are causally connected:


Method of Agreement  If two or more instances of the phenomenon under investigation have only one circumstance in common, the circumstance in which alone all the instances agree is the cause (or effect) of the given phenomenon.


Method of Difference If an instance in which the phenomenon under investigation occurs, and an instance in which it does not occur, have every circumstance in common save one, that one occurring in the former, the circumstance in which alone the two instances differ is the effect, or the cause, or an indispensable part of the cause, of the phenomenon. 


Method of Concomitant Variation 

The magnitude of a phenomenon P is a function of factors A, B, C....

 The set of factors on the right hand side of this equation is the full cause of P, while each of the relevant factors is a partial cause. 


                                (System of Logic; Book III, Chs 1-8)


John Mackie (1917-1981)


1.             Introduces the idea of a field:

·         This meets the problem that the causes of a simple event like the matchhead igniting seem on the face of it to be infinite in number.

·         In Mackie’s analysis the background conditions and the counterfactual conditions, are set aside as part of the field.  This also solves familiar problems such as:   

‘Since it is normal for people to be striking matches and lighting cigarettes in a residential flat, but a gas leak is abnormal and should not occur, we may well say that the explosion which wrecked this block of flats was caused by the presence of a quantity of gas rather than that it was caused by Jones lighting his cigarette’. 



2.             INUS Cause


·         Frequently it is not only the case that (Mackie pp 61-2)

‘All ABC are followed by P’

 but also

‘All DGH are followed by P’; 

·         It could be that


‘All ABC and/or All DGH and/or All JKL are followed by P’;

‘All P are preceded by  ABC or  DGH or  JKL or ABCDGH or ABCJKL or DGHJKL or ABCDGHJKL’


‘X is a necessary condition for Y’means when an event of type Y occurs an event of type X also occurs;

X is a sufficient condition for Y’ means when an event of type X occurs an event of type Y also occurs; 


·         Then

 [ABC or  DGH or JKL] - is a condition which is both necessary and sufficient for P;


·         and

[ABC] is a condition which is sufficient but not necessary for P

[DGH] is a condition which is sufficient but not necessary for P

[JKL] is a condition which is sufficient but not necessary for P;

[ABCDGH] is a condition which is sufficient but not necessary for P

[ABCJKL] is a condition which is sufficient but not necessary for P

[DGHJKL] is a condition which is sufficient but not necessary for P

[ABCDGHJKL] is a condition which is sufficient but not necessary for P


·         Each single factor (A, B, C, D, G, H, J, K, L) is neither necessary nor sufficient for P; 

·         But each is clearly related to P in a significant way:

                A (...N)  is an Insufficient but Non-redundant part of an Unnecessary but Sufficient condition for P




Wesley Salmon (b.1925)


Two kinds of causal mechanism:


1) spatiotemporally continuous causal processes that transmit causal influence from one part of spacetime to another.   (Note the term ‘process’).

2) causal interaction: when two or more causal processes intersect in spacetime they may or may not produce lasting modifications in one another


These in turn enable us to sort out two aspects of causal explanation:


1) aetiological explanation: tells the story of how one event caused another and so on – a caribou died 30,000 years ago, its carcass was preserved in ice for many millenia, later  (5,000 years ago) a  bone was found by a human artisan, s/he carved it into a useful and decorative object.

2) constitutive explanation: the pressure exerted by a gas on the walls of a container is explained in terms of momentum changes between the molecules and the walls (this is like temperature and ice, bricks and glass, branches and cradles, sugar and syrup).


3) Salmon also wants to solve a problem from probability that we have met before. 

·         Hempel argues that we have an explanation when we have a set of conditions and probabilistic ‘laws’ such that the combination of conditions and laws means that there is a high probability that the explanandum would occur.

·         However, supposing it were the case that overall it turns out that 20% of HIV positive people contract AIDS.  Are we less likely to say that HIV causes AIDS than in the alternative case that 80% of HIV positive people contract AIDS?  In either case HIV looks like an INUS cause of AIDS. 

·          And Salmon would want to say that we know HOW HIV causes AIDS (attacking the immune system, causing elements of the immune system to misrecognise cells, .... ) so there is more to the relationship between HIV and AIDS than constant conjunction or regularity, as well as more to it than prediction.   


‘On the view of causality  I am advocating, causal connections exist in the physical world and can be discovered by empirical investigation’. (p23).


Salmon’s philsophy of causality

‘involves relinquishing rational expectability as a hallmark of successful scientific explanation.  Instead of asking whether we have found reasons to have expected the event to be explained if the explanatory information had been available in advance, we focus on the question of physical mechanisms.  Scientific understanding, according to this conception, involves laying bare the mechanisms – aetiological or constitutive, causal or noncausal – that bring about the fact to be explained.  If there is a stochastic process that produces one outcome with a high probability and another with low probability, then we have an explanation of either outcome when we cite the stochastic process and the fact that it gives rise to the outcome at hand in a certain percentage of cases.  The same circumstance – the fact that this particular stochastic process was operating – explains the one outcome on one occasion and an alternative on another occasion.’.  (p328)



Bas van Fraassen (b.1941)


Against philosophers like Mackie and Salmon van F argues, following Hanson:

There are as many causes of x as there are explanations of x. Consider how the cause of death might hav ebeen set out by a physician as ‘multiple haemorrhage’, by the barrister as ‘negligence on the part of the driver’, by a carriage builder as a ‘defect in the brakeblock construction’, by a civic planner as ‘the presence of tall shrubbery at the turning’.  N R Hanson Patterns of Discovery 1958 p54




example  The judge asks Robber Robert why he robbed the bank.  RR answers: ‘because that’s where the money was.’ 

                Judge:    why rob the bank (as opposed to live a law abiding life)?

                RR:          why rob the bank (as opposed to the chemists)?



‘It is not the case that the meaning of the sentence ‘A is the cause of B’ depends on the nature of the phenomena A and B, but that this meaning depends on the context in which this sentence is uttered.  The nature of A and B will in most cases play a role, indirectly, but it is in the first place the orientation or the chosen point of view of the speaker that determines what the word cause is used to signify.’