Counterfactual Conditionals Author(s): Lauri Karttunen Reviewed work(s): Source: Linguistic Inquiry, Vol. 2, No. 4 (Autumn, 1971), pp.

566-569 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/4177665 . Accessed: 08/04/2012 10:53
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Linguistic Inquiry.

http://www.jstor.org

566

SQUIBS

AND

DISCUSSION

inference associated with since (by virtue of its temporal meaning) to a piece of the semantic content of since.*

COUNTERFACTUAL CONDITIONALS

LauriKarttunen, University of Texasat Austin

It has often been assumed that a counterfactual conditional presupposes the negation of both the antecedent and the consequent.1 For example, Lakoff (I970) takes it to be a fact that (i) presupposes both (2a) and (2b). (i) If Harry had known that Sheila survived, he would have gone home. (2) a. Harry did not know that Sheila survived. b. Harry did not go home. It seems to me that this analysis is mistaken. I will try to show that (2b) is not presupposed by (i). Of course, I do not deny that, in most contexts, (i) suggests very strongly that Harry did not go home. Nevertheless, it does not necessarily follow from this that (2b) is presupposed. There is an alternative, better analysis. The crucial argument against regarding (2b) as a presupposition is pointed out by Lakoff himself. Without any contradiction, the speaker may continue the sentence in (i) in a way which entails that the consequent is actually true and (2b) false. For example, there is nothing contra-

dictoryabout (3).
(3) If Harry had known that Sheila survived, he would havc gone home, which he did anyway.

Similarly, (4) is perfectly consistent, although the word still indicates that the consequent is to be taken as true. (4) If Harry had known that Sheila survived, he would still have gone home.

* This work was supported in part by the 1970 MSSB Advanced Research Seminar in Mathematical Linguistics, sponsored by the National Science Foundation through a grant to the Center for Advanced Study in the Behavioral Sciences, Stanford, and held at the Ohio State University. We are indebted to Lauri Karttunen for many insightful comments. 1 This assumption is trivially false unless we exempt sentences of the following type. (i) If Fran had baked a pie, she would have eaten it. (ii) If Fran had eaten a pie, she would have regretted it. In both of these examples, the consequent can be true (or false) only in a context where the antecedent is true. In other words, the consequent presupposes the antecedent. Furthermore, the antecedent is presupposed to be false in the actual world. It follows that, in the actual world, the consequent can be neither true nor false. In the following, I will limit myself to cases where it is possible for the antecedent and the consequent to be false simultaneously.

SQUIBS

AND

DISCUSSION

In Lakoff's view, (3) and (4) nevertheless presuppose the negation of the consequent. It just happens that they also contain a "qualifying phrase". The phrase which he did anywayin (3) and the word still in (4) act as qualifying phrases which "cancel" this presupposition. I find Lakoff's notion of "cancelled presupposition" unintelligible and will comment on it later. At this point, it is important to notice that the speaker cannot similarly indicate, without contradicting himself, that he regards the antecedent clause of a counterfactual conditional as true.2 A sentence like (5) is inconsistent, if not incoherent. (5) *If Harry had known, as he did, that Sheila survived, he would have gone home. As I see it, although a counterfactual conditional presupposes the negation of the antecedent, it only suggests, in the absence of a disclaimer, that the consequent is false too. It seems to me that the logical form of (i) is something like (6): (6) a. Presupposition: Harry did not know that Sheila survived. If Harry knew that Sheila b. Assertion: survived, Harry went home. This analysis of (i) does not itself explain why that sentence so strongly suggests that Harry did not go home: however, it constitutes a part of the explanation. What is needed in addition is a general principle discussed by Geis and Zwicky in the accompanying squib. As Geis and Zwicky point out, there is a natural tendency in the human mind to perfect conditionals to biconditionals. Students in an elementary logic course often propose that examples such as (7) are to be formalized as biconditionals rather than conditionals. (7) If you mow the lawn, I'll pay you five dollars. That is, most people feel that the appropriate logical form of statements like (7) is the conjunction of (8a) and (8b). (8) a. S 1S2 b. S1 D S2 This is not quite right since (8a) alone is enough. However,
2 It has been suggested to me (J. R. Ross, personal communication) that the difference can be accounted for by regarding the negation of the antecedent as a "second order presupposition". Namely, according to Lakoff only "first order presuppositions" can be cancelled. However, Lakoff admits (1970, 46) that there also are several types of first order presuppositions which cannot be cancelled. For this reason, I do not see that we have any basis for deciding whether the negation of the antecedent is a first order or a second order presupposition, Lakoff himself (1970, 38) puts it into the former class,

568

SQUIBS

AND

DISCUSSION

it is obvious that in a great majority of cases where a conditional like (7) is uttered, (8b) is also tacitly assumed. In natural language, (7) suggests rather strongly that, if you don't mow the lawn, I won't pay you five dollars. What would be the point of stating a condition if it was not a necessary condition? According to the principle proposed by Geis and Zwicky, any assertion of the form (8a) suggests, or "invites the inference", that the corresponding assertion of the form (8b) is also true. In the case of our semantic analysis of (i), the Geis-Zwicky principle predicts that the assertion (6b) invites the inference (6b) below. (6b) *If Harry did not know that Sheila survived, Harry did not go home. Assuming that the principle is correct, we have an explanation for the fact that (i) is likely to be understood to mean that Harry did not go home. This follows directly from the conjunction of the presupposition, (6a), with the invited inference, (6b), by Modus Ponens, as shown in (9). (9) (6a) Harry did not know that Sheila survived. (6b) *If Harry did not know that Sheila survived, Harry did not go home. .,. (2b) Harry did not go home. Note in particular that one of the two premises in (9), namely (6b), is only an "invited inference"; it is neither asserted nor presupposed by (i). Consequently, the speaker is not in any way committed to the view that Harry did not go home. He may indicate that the consequent is true, as we have seen in (3) and (4), without making himself guilty of inconsistency. Examples like these do not show that a presupposition can be "cancelled" by a qualifying phrase, as Lakoff (1970) mistakenly assumes. It is only the antecedent clause that is presupposed to be false in a counterfactual conditional. Of course, I do not wish to claim that invited inferences are involved in all cases where Lakoff sees a presupposition that has been "cancelled out". On the contrary, it seems to me that the two types of examples that he discusses in this connection do not really belong together. Consider the following sentence: (io) John has stopped beating his wife, if heeverbeat herat all. According to Lakoff, the qualifying phrase if he everbeather at all cancels out (i I), which otherwise would be presupposed by (io).
(I I)

John has beatenhis wife.

(i

To me it seems that the presence of the if-clause in

o)

SQUIBS

AND

DISCUSSION

indicates that the speaker is not sure whether (i I) is true. The assertion "John has stopped beating his wife" is made only conditionally.3 Compare this case with the example in (3), where the phrase whichhe did anywayasserts the truth of the preceding consequent clause. If we are to believe Lakoff, in (3) the truth of a presupposition is denied. On the other hand, in (i o) one cannot go as far as to deny the truth of (i i). The following example is anomalous: (I2) *John has stopped beating his wife, and he never beatherat all. Since Lakoff does not distinguish between these two cases, I fail to understand what "cancelling out a presupposition" is supposed to mean. I do not believe that (3) and (io) have anything in common which would justify grouping them together as examples of the same phenomenon.* References Geis, M. and A. M. Zwicky

(I97I)

"On Invited In561-566.

Inquiry, ferences," Linguistic 2.4, pp.

Lakoff, G. (1970) Linguisticsand Natural Logic, Studies in Generative Semantics, No. i, Phonetics Laboratory, The University of Michigan. (To appear in Synthetse.) Ross, J. R. (I970) "On Declarative Sentences," in R. A. Jacobs and P. S. Rosenbaum, eds., Readings in Grammar,Ginn and Co., English Transformational Waltham, Mass.

MIRROR-IMAGE

RULES AND

VSO

ORDER*

Ross, John Robert Foundation Research Language and MIT

In Langacker (I969), Langacker pointed out the existence of syntactic processes which operate in two directions. While I suspect that some of the rules he cited as examples of this phenomenon (Gapping, for instance) may turn out to be reanalyzable in other terms, it is undeniable that some rules which have this property exist. In particular, it seems
* This work was supported by the 1970 MSSB Mathematical Linguistics Seminar, sponsored by the National Science Foundation through a grant to the Center for Advanced Study in the Behavioral Sciences, Stanford, and held at The Ohio State University. 3 If one accepts J. R. Ross's (I970) proposal about underlying performative sentences, one may want to derive (io) from a structure such as (i): (i) If John has ever beaten his wife at all, then I asserttoyou that John has stopped beating his wife. Somehow it has to be made explicit that (i o) is no ordinary conditional. * This research was supported in part by a grant from the National Science Foundation (Number GS-3202).