[Lukes, Steven (1980), Elster on counterfactuals (Review of Logic and Society), Inquiry 23 (2):144-155]

[ Books | Articles | Reviews | Index | The main menu ]

Elster on Counterfactuals

Steven Lukes

Balliol College, Oxford

[start of page 145]

Jon Elster's chapter 'Counterfactuals and the New Economic History' in his Logic and Society is unusually stimulating and it displays to the full the characteristic virtues of his approach ofwhich two are pre-eminent. First, his treatment of issues in the philosophy of sodal science and of history genuinely engages with problems of and disputes among practitioners. He sees that counterfactual reasoning is not merely a technical problem for logicians and philosopliers but a live issue among historians who in this respect are for the most part like M. Jourdain speaking prose. For such reasoning, as Elster recognizes, is an indispensable component of all explanatory theorizing. Max Weber, of course, observed this long ago, remarking that history (meaning the writing of history)

does recognise possibilities, assuming that it seeks to be a science. In every line of every historical work, indeed in any selection of archival and source materials for publication, there are, or more correctly must be, 'judgments of possibility', if the publication is to have value for knowledge.1

[end of page 145, start of page 146]

Elster puts the point in a more modern idiom, to which we will return:

[P]ossible worlds are a hidden and implicit aspect of all model-building and of all theorizing . . . A theory must have implications for possible worlds by specifying the set of jointly realizable values of the relevant variables.2
What is particularly valuable is Elster's attempt to determine the conditions for successful counterfactual reasoning by working through the real-life explanatory arguments and disputes of contemporary economic historians.

In the second place, I applaud Elster's ambition to show that 'logical theory can be applied not only in the formalization of knowledge already obtained by other means, but that logic can enter in the creative and constructive phase of scientific work' .3 His analysis of countetftctual reasoning is a serious attempt to contribute to ground-floor (as opposed to meta-) explanatory success. He presents logicians' analyses of counterfactuals (particularly that of David Lewis) and of the criteria of identity (particularly that of Kripke and Dummett) and tries to show how these might bear on his chosen examples of historical reasoning. He develops his own account of what makes counterfactuals (a) legitimate and (b) assertable, claiming superiority for it as against that of Lewis, on philosophical but mainly pragmatic grounds: that is, if he is right, then practising historians and social scientists, by following his recommendations, will both be clearer about what they actually do and produce better explanations.

What I propose to do here is to suggest that, despite these considerable virtues, Elster's approach, as put to work here, has misfired in a number of ways. I shall argue, first, that his classification of the various approaches to the problem among logicians and philosophers is inadequate and confusing, and that this lends a certain confusion to his own treatment of the subject; and that as a result, secondly, he makes exaggerated claims for his own theory, which itself only relocates the central problem, and, thirdly, he is led into a mistaken account of when counterfactuals are illegitimate.

Elster begins by drawing an over-all distinction between meta-linguistic and ontological approaches to counterfactuals, the former making statements about the world rather than non-linguistic features of the world 'the crucial explanatory element', the latter claiming that counterfactuals are

[end of page 146, start of page 147]

about possible worlds, and are rendered true or false by the features of these worlds.4 He himself adopts, he says, the first approach, which he identifies with the position that 'in historical writings (and possibly in general) counterfactual statements should always be understood in the context of an implicit or explicit theory '5 But there is a third, distinct approach to counterfactuals which treats counterfactuals neither as second-order linguistic performances (endorsing certain combinations ofantecedent and consequent as true or assertable) no as statements rendered true or false by features of a world, whether possible or actual. I refer to the so-called 'condensed argument' and closely related 'suppositionist' theories, according to which counterfactuals are, respectively, either abbreviations of arguments, with auxiliary premisses omitted or unformulated, or suppositions, where the consequent is asserted within the scope of the supposition of the antecedent, which is taken to be unfulfilled.6 Here the counterfactual is the (telescoped) argument or supposition itself, not a higher-level statement about these, but it is not strictly capable of truth or falsity, though it is capable of plausibility, validity, etc. Elster does, it is true (following as he says, a paraphrased quotation from Lewis), appear to refer to the condensed argument approach, and indeed to adopt it, but, confusingly, he describes it as a variant of the meta-linguistic approach.7 But the latter label properly applies to an interpretation of counteffactuals as stating that the counterfactual conditional 'if P, Q' is equivalent to 'either "P" entails "Q" or there is some statement or set of statements "S" which is true, and cotenable with "P", and the conjunction of which with "P" entails "Q" '.('cotenability' then being in need of definition).8 The condensed argument approach, by contrast, asserts that a counterfactual is just that: an argument, about the world.

Now, oddly, Elster spends a good deal of time trying to establish a better criterion than that furnished by Lewis for assessing the distance between worlds and, in particular, the closest world to the actual world. He has some very telling objections to Lewis, in particular that his notion of similarity does not allow for unambiguous ordering, that his invocation of the notion of 'importance' is circular, since it rests on causal judgments (which the analysis of counterfactuals purports to elucidate), and that Lewis seems to favour singling out as the closest possible world a world that could not feasibly have come about (though it is contestable whether this last objection is fair or indeed an objection). But in following Lewis and appealing to possible worlds and the distance between them, Elster generates confusion, since the rationale for that approach is that it prom-

[end of page 147, start of page 148]

ises to supply counteifactuals with truth conditions to give them determinate meaning and truth values, independent of their users and their purposes. Yet Elster is also very clear that possible worlds are 'just facon de parler that can always in principle be dispensed with. You cannot get more out of them than you have put into them yourself';9 and he insists that counterfactuals are theory-using, not theory-confirming. It seems to me that his talk of 'possible worlds' is at odds with his recognition that the choice of what elements of the actual world are to be retained when a belief-contravening supposition is entertained, is theory-dependent and is in that sense relative to the theorist and his explanatory purposes.

In short, Elster claims to follow the meta-linguistic approach, uses the idiom of the ontological-possible-worlds approach but would, so far as I can see, be better advised in terms of his own intuitions and purposes to adopt, unequivocally, the condensed argument approach. As I shall try to show, this would not only make his account clearer and less confusing: it would also improve it.

Elster's own positive contribution to the analysis of counterfactuals is to offer what he calls a 'branching worlds theory',10 according to which a historian's counterfactual assertion 'is intended as, and must be analysed as, a statement about what could have happened (for all that we believe) to the real past'.11 More particularly, the 'counterfactual antecedent must be capable of insertion into the real past':12 there must be a permitted trajectory from a branching point in the actual history of the system under investigation which leads to the state of affairs it describes. Following Lewis, Elster holds that the counterfactualizing historian's task is to identify the closest possible world; he identifies the closest world where the antecedent obtains by seeking the latest point in the real past which leads, by a feasible path, to the state described by the antecedent, and then inquires whether the consequent also holds. The further back one has to go in history to insert the possible state into real history, the greater the distance to that state. Thus for Elster historical counterfactuals are assertable if the consequent holds in the closest world(s) where the antecedent obtains. So for instance, the statement 'If it had not been for slavery, the GNP of the US South in 1860 would have been twice as high as it actually was' would not be assertable 'if a non-slave South could stem from a branching point no later than, say, 1750, whereas a GNP of the

[end of page 148, start of page 149]

required size would require counterfactual changes going back to 1700'. 13 Or again, in seeking to assess the impact of the introduction of the railroad on the American GNP of 1890, Fogel, according to Elster, rightly goes back to about 1830 where 'one could legitimately assume a branching point without railroads',14 and explores what alternative developments would have occurred in the period up to 1890 though Elster also criticizes Fogel for being inconsistent and reasoning, to some extent, as though the appropriate procedure is to take the actual 1890 economy and remove all features directly linked to railroads.

This is a valuable discussion and it certainly helps in distinguishing between different forms of counterfactual speculation: in particular, between thought-experiments about the consequences of the non-introduction of (say) slavery or the railroads and those about the consequences of their removal. He may, further, be right to suggest that, broadly speaking, historians tend to be properly more interested in the former kind of speculation than the latter: they are naturally interested in 'feasible paths', while other kinds of thought-experiment leave them cold. It may well be, for instance, that a 'possible world' constructed by taking the actual 1860 economy and eliminating all features of society that stem directly from the present institution of slavery would be 'extremely uninteresting if we are asking about the importance of slavery for Southern economic growth'.15

On the other hand, Elster has not shown that his account is, in any clear sense, the 'correct' account of counterfactuals, or even of historical counterfactuals, and indeed his requirement that 'a counterfactual antecedent must be capable of insertion into the real past', by derivation from a branching point, as a condition of the legitimacy of the counterfactual, seems patently excessive. For, even if historians may find some counter-factual suppositions which are incapable of such insertion 'extremely uninteresting', that does not show them to be 'illegitimate', except in a very weak sense.

But in fact many counterfactuals that do not obviously satisfy Elster's criterion for legitimacy are highly interesting and, implicitly or explicitly, appealed to by practising social scientists and historians. We might, for example, hold that Trotsky could never have been in Stalin's position, for a variety of personal and political reasons, and still be intensely interested in what difference there would have been between his rule and Stalin's. Equally, we might well be intensely interested in the hypothetical consequences of Stalin's removal, together with those features stemming directly from his autocracy, in, say 1929, for the functioning of the Soviet

[end of page 149, start of page 150]

State, as opposed to the consequences for that St~-tte of an alternative non-Stalinist path. Historians are not only interested in what other developments were historically possible; they are also interested in assessing the difference the particular factors made to a situation during a certain period, as opposed to the difference they made to the generation of that situation. This involves 'thinking away' the isolable contribution of those factors within some specified time-period. Of course, the consequences attributable to such a factor must, indeed, be isolable: as Elster well shows, John Stuart Mill, for one, thought that this could never be so in social life. But assuming (with Elster) that Mill was wrong about this, there seems to be nothing illegitimate about thought-experiments such as the following classic example from Saint-Simon:

Suppose that France preserves all the men of genius that she possesses in the sciences, fine arts and professions, but has the misfortune to lose in the same day Monsieur the King's brother, Monseigneur Ie duc d'Angouleme, Monseigneur le duc de Berry, Monseigneur Ie duc d'Orleans, Monseigneur Ie duc de Bourhon, Madame Ia duchesse d'Angouleme, Madame Ia duchesse de Bourbon, and Mademoiselle de Conde. Suppose that France loses at the same time all the great officers of the royal household, all the ministers (with or without portfolio), all the councillors of state, all the chief magistrates, marshals, cardinals, archbishops, bishops, vicars-general, and canons, all the prefects and sub-prefects, all the civil servants, and judges, and, in addition, ten thousand of the richest proprietors who live in the style of nobles.

This mischance would certainly distress the French, because they are kind-hearted, and could not see with indifference the sudden disappearance of such a large number of their compatriots. But this loss of thirty-housand individuals, considered to be the most important in the State, would only grieve them for purely sentimental reasons and would result in no political evil for the State.16

If Elster rules out too many quite respectable and interesting counterfactuals as 'illegitimate', he is also, in the end, less than illuminating about those he rules in as 'legitimate' and 'assertable'. For he gives us no real help in answering the next question his account inevitably raises: namely, how do we identify branching points? This, of course, is a highly theory-dependent matter. It is obviously a matter for speculation and argument as to whether a given point in historical time was or was not at the origin of an alternative feasible path to the actual course of history. Germany may have been at such a point in 1918: a moderate liberal path in Weimar Germany might have been possible from that point. To make such a claim is, precisely, to advance argument, to adduce reasons that may be more or less compelling and indirect evidence that may be more or less relevant, in the context of a web of suppositions, claimed to be plausible.17 Incidentally, Elster makes a general suggestion with respect to such claims

[end of page 150, start of page 151]

that I find implausible: namely, that 'the subjective range of possibilities is a natural joint in which to insert the wedge of counterfactualization'.18

The plausibility of such arguments is precisely what is at issue in discussing the logic of counterfactuals, and yet Elster writes as though, once the branching point has been identified, the assertability of the counterfactual can be decided by examining the compatibility of the antecedent and the consequent. But this begs the question, since everything hinges on the identification of the branching point. Once more, I suspect that Elster's temptation to use the idiom of 'possible worlds' leads him away from his half-appreciated insight that counterfactuals are condensed arguments. Elster's way of talking strongly suggests that there is a discoverable determinate answer, a 'fact of the matter', to the question of where the branching point lies, whereas the historian is engaged in supposing some but not all features of the world to have been other than they were with a view to assessing certain consequences of that supposition. Both the supposition and the assessment are creatures of his theory, though insofar as his thought-experiment retains unknown contingent features of the actual world and the continuing operation of its causal laws, constraints are set on possible outcomes that are independent of his imaginings.

In the third place, it seems clear that Elster makes a mistake in his presentation of what he calls the 'scissors problem'. Theories, he claims, play a double role: as 'filter for the acceptance or the rejection of the antecedent' and as deductive machinery, enabling us 'to conclude from the hypothetical antecedent to the hypothetical consequent' 19 His argument is that 'the nomological laws that we use in concluding from the antecedent to the consequent of a counterfactual statement must also be used in order to ensure that the antecedent itself is compatible with the elements that are not assumed to vary' 20 This alleged situation leads to the following dilemma:

[T]he stronger (i.e. the more deterministic) the theory T, the better grounded is the conclusion from antecedent to consequent, but the more vulnerable is also the legitimacy of the antecedent. With a weak theory many antecedents are permitted by the filter, but it may be impossible to prove the assertability of the conditional. Thus for a successful counterfactual analysis a delicate balance must be struck: the theory must be weak enough to admit the counterfactual assumption, and also strong enough to permit a

[end of page 151, start of page 152]

clear-cut conclusion . . . when evaluating the consequences of a non-railroad economy in 19th-century America, we can hardly ask whether the internal combustion engine would have been invented before it actually was, for this would require a theory of technical change that might prevent the non-railroad assumption itself from being meaningful. The problem is not that it is difficult to know whether the internal combostion engine would have been invented, but rather that if it were possible to answer this question, it should not have been put in the first place.21

If we impose some dynamic condition of legitimacy, we know that the emergence of more and better theories will ultimately reduce the set of assertable counterfactuals, for in the limit - an all-embracing and deterministic theory of the universe - no branching points will exist where the wedge of counterfactualization can be inserted.22
To this argument it has been objected, by Brian Barry,23 that Elster is quite unjustified in distinguishing in this way between legitimate and illegitimate counterfactuals. Barry simply asks why we should not be permitted to imagine what our theories show to be impossible. Do we not do it all the time? Since (as Elster admits) counterfactuals are merely devices for talking about the implications of our theories, there is no need to worry about how we get from the real world to the initial conditions posited by the counterfactual. We might be absolutely sure that the world could not have satisfied the antecedent of a counterfactual, but 'that does not prevent the counterfactual from being a useful way of drawing attention to the implications of what we do believe' 24 Thus, with regard to the railroad, Barry argues that if 'we want to find out what would have happened without railways . . . then we should follow out our theories of economic growth, invention and whatnot without railways. There is no need to specify why there were no railways; we can simply postulate it' 25

To this objection Elster replies26 that his theory specifically rules out only 'the antecedent whose impossibility is determined by the very same knowledge that we use in order to derive the consequent' 27 Thus, with regard to the railways, he states his belief that

we cannot in one breath assume that our knowledge about technical change is so weak that we may assume the non-invention of the railroad without getting into conflict with our theories, and then in the next breath assume that we know so much that we are able to predict the substitutions that would have developed. Max Weber says somewhere that historical materialism is not like a taxi one may enter or leave at will; and the same holds, I believe, for theories in this context.28
I believe that Elster has been led astray here and that Barry is basically right. The supposition of no railways does not imply that our knowledge of

[end of page 152, start of page 153]

technical change is weak -or indeed that our theory is non-deterministic. It is, rather, an explicitly belief-contravening supposition, introduced into an argument for a clear purpose - namely, to try to illuminate, on the basis of the theories we hold, what difference the introduction of the railways made. Elster's suggestion that if 'we accept a strong theory of technical change and stipulate the non-invention of the railroad by, say, 1835, then our theory would lead us to re-invent them by 1836' 29 is weak, since, though plausible, it does not demonstrate the illegitimacy of supposing the non-invention of the railroad during the entire period under historical consideration.

More generally, it appears to me quite arbitrary to propose that counterfactual reasoning should be confined to imagining (a subclass of) possible antecedents - namely, those which the theory we use to infer the consequent tells us are compatible with the elements of the world that are not assumed to vary. Why should we not be allowed to imagine the continuing non-existence of the railroad - even though we think we know that it had to be invented - in order to see what other transport technologies and what forms of distribution would have been established? Why should we not imagine Trotsky in Stalin's place, even though our theory of Trotsky, Stalin, and the relations between them tells us that he could never have made it? Why, in general, should we not be allowed to violate laws we otherwise accept, in order to bring out their implications'? What else is the meaning of the philosophers' phrase 'per impossibile'? Is it to be banned from the lexicon?

I think that Elster has been misled by his temptation to think that historical counterfactuals must be about real possibilities in the past.30 For this temptation, allied with his talk of possible worlds, tends to deflect him from perceiving the suppositional character of counterfactuals and their status as (telescoped) arguments. He has, in fact, much to say concerning the ways in which counterfactuals, thus understood, can fail - if, for example, there is a failure to preserve identity from antecedent to consequent, or if the consequent of the counterfactual refers to an agent whose identity is conceptually linked to features of the actual world whose removal is postulated by the antecedent. Doubtless many antecedents may be declared illegitimate in ways like this - for example, if they require us to make contradictory suppositions.31 But what Elster has not demonstrated is that we are not permitted to suppose, in contravention of our actual beliefs, that the laws we accept are suspended in some specified sphere, but otherwise applicable.

[end of page 153, start of page 154]

1 Max Weber, 'Objective Possibility and Adequate Causation in Historical Explanation' in The Methodology of the Social Sciences, trans. and ed. by E. A. Shils and H. A. Finch, Free Press, New York 1949, p.173.

2 J. Elster, Logic and Society: Contradictions and Possible Worlds, Wiley, Chichester/New York/Brisbane/Toronto 1978, p.7.

3 Ibid., p.1.

4 Ibid., p.181.

5 Ibid., p.5, original emphasis.

6 For a systematic discussion of approaches to interpreting conditional statements, incltiding counterfactuals, see i. L. Mackie, Truth, Probability and Paradox (Clarendon, Oxford 1973), Ch. 3. Mackie defends the suppositionist account.

7 Logic and Society, p. 182. He writes that 'the crucial point is that the counterfactual statements must be explained in terms of some actual theory T that the speaker, if challenged, could produce to back his assertion' (ibid.).

8 See Mackie, op. cit.

9 Logic and Society, pp. 7-8.

10 Ibid., p.180.

11 Ibid., p.218.

12 Ibid., p.184.

13 Ibid., p.191.

14 Ibid., p.204.

15 Ibid., p.211.

16 F. M. H. Markham (Ed.), Henri Conite de Saint-Simon (1760-1825): Selected Writings. Basil Blackwell, Oxford 1952, pp. 72-73.

17 For such an argument, concerning the case cited, see Barrington Moore (Jr.), Injustice: The Social Base's of Obedience and Revolt, Macmillan, London 1978. Ch. II, 'The Suppression of Historical Alternatives: Germany, 1918-20' explores the thesis that in Germany immediately after the First World War, 'conditions may have existed that rendered possible some sort of a liberal breakthrough resulting in a regime more stable than the one that in a decade and a half was to succumb to Adolf Hitler' (p.379).

18 Logic and Society, p.180. Such a suggestion prejudges the extent to which unintended consequences are in operation, so that history works behind men's backs. Interestingly, Max Weber, to whom Elster appeals in making this suggestion, himself observes that the historian may be in a better position than Bismarck to assess the anticipatable consequences of alternative decisions ('Objective Possibility and Adequate Causation. . . , p. 165).

19 Logic and Society, p.184.

20 Ibid., p.180.

21 Ibid., pp. 184-85

22 Ibid., p.186.

23 Brian Barry, 'Superfox', Political Studies, Vol. VIII (1980), No.1, pp. 138-43.

24 Ibid., p.139.

25 Ibid., p.141.

26 'The Treatment of Counterfactuals: Reply to Brian Barry', Political Studies, Vol. XVIII (1980), No.1, pp. 144-7.

27 Ibid., p.145.

28 Ibid., p.146

29 Ibid., p. 146

30 It is unclear if he confines this doctrine to historians' counterfactuals. It is noteworthy that in his reply to Barry, he cites as an example falling under his account the theory of evolution as applied to kangaroos.

[end of page 154, start of page 155]

31 This is surely what explains Elster's discomfort with his de Gaulle example (Logic and Society, p. 185): to imagine a de Gaulle in power and capable of sacrificing France's independence is to entertain incompatible suppositions about de Gaulle. The same applies to Barry's example of Julius Caesar: the supposition that Caesar had crossed the Rubicon is incompatible with the supposition that he is a cautious man.

[end of page 155]

[Lukes, Steven (1980), Elster on counterfactuals (Review of Logic and Society), Inquiry 23 (2):144-155]

[ Books | Articles | Reviews | Index | The main menu ]

1 1