Approaching double fake tensed conditionals Katrin Schulz In this talk we will focus on a particular problem concerning the interpretation of tense in conditionals: Fake Tense. Fake Tense refers to the observation that in English subjunctive conditionals1 the Simple Past, and sometimes also the Past Perfect, appear not to be interpreted as semantic past tense or past perfect. For illustration, in the antecedent of the indicative conditional (1.a) the finite verb is marked for the Simple Past. As expected, the antecedent refers to a past leaving of Peter. Example (1.b) is a simple past subjunctive conditional (SP-C) with exactly the same antecedent, but now Peter’s leaving has to take place in the future. The past perfect subjunctive conditional (1.c) (PP-C) also refers to the future but has even two layers of past morphology. This raises questions about the semantic function of the past tense markers in these sentences. This talk will be focussing on examples like (1.c) were Simple Past as well as the Perfect appear to be fake. (1) a. b. c. If Peter left in time, he will be in Frankfurt this evening. If Peter left in time, he would be in Frankfurt this evening. If you had been in Paris next week, we could have met. Two different lines of approach to fake tense can be distinguished in the literature: pastas-modal approaches (PaM) claim that the past tense markers receive in these contexts a modal interpretation; past-as-past approaches (PaP) propose that the past still receives a temporal interpretation, though it contributes in an unexpected way to the meaning of the sentences. In this talk I will argue against a PaM treatment of the second layer of fake past morphology in examples like (1.c). Instead, I will propose a PaP approach to the Perfect in (1.c). This approach will build on an interventionist account of counterfactuals using causal structural models ([1], [2]). There is very little discussion of the second layer of fake past in conditionals like (1.c) in the literature. An exception is the PaM approach of [6], which builds on [2]. [6] proposes that the past tense morphology in subjunctive conditionals carries a presupposition of distance from the epistemic deictic center, as much as it presupposes distance from the temporal deictic center in its temporal interpretation. Read as modal the Simple Past marks that the described eventuality is at least unexpected according to the beliefs of the speaker. This approach can be extended to the PP-Cs. A modal interpretation of the double past in PP-Cs conveys that the expressed proposition is false. This accounts for sentences like (1.c) for which counterfactuality is reported to be not cancellable. This proposal is challenged by generic counterfactuals (If kangaroos had no tails, they would topple over). I will present two empirical studies showing that there is a strong preference to use SP-Cs to express these conditionals, even in case the counterfactuality is common knowledge and relevant. Given the principle Maximize Presupposition, [6] seems to predict that in such situations a PP-C should be preferred over a SP-C, since it comes with the stronger presupposition that the conditional is counterfactual (not just unexpected). If the observed preference of a SP form over a PP form would apply to all counterfactuals, the approach could be defended by pointing out that the PP form is more complex than a SP-C and, therefore, the principle Maximise presupposition doesn’t apply. However, this doesn’t seem to be the case. Already in [3] it has been observed that in certain contexts the PP form is preferred. If in the past it has already been decided that the antecedent will not become true2 , then the speaker needs to use a PP-C. This 1 2 This holds also for many other languages. In the examples of [3] the event described in the antecedent already happened in the past. 1 cannot be explained by the PaM approach to PP-Cs sketched above. I propose that we should opt for a PaP approach, instead. Most PaP-approaches adopt the idea that conditionals are modal construction. The relevant modal base for subjunctive conditionals is generally taken to be the set of alternative futures for the actual world at some time t. According to PaP-approaches the function of the Simple Past in subjunctive conditionals is to move the modal base to a past time. As an approach to SP-Cs the approach is problematic exactly because of the observation that for generic counterfactuals the SP form is preferred. Moving the modal base back in time doesn’t make sense for these counterfactuals. However, for the same reason a PaP approach to the perfect in PP-C could explain the observation that the PP form is barely used for generic counterfactuals. The proposal. I propose that we keep a PaM-approach for the Simple Past in subjunctive conditionals, but adopt a PaP-approach for the second layer of past tense markers in PP-Cs. The proposal is is based on a interventionist approach to the meaning of counterfactual conditionals ([1], [4], [5]). We add a causal network ([4]) to our model to keep track of the relevant (causal) laws/generalisations. Conditional are evaluated as follows: we first change the cause of events using an intervention. An intervention is a manipulation of the world that forces the antecedent to be true. Then we use the information on the laws to infer which changes follow from the intervention. Finally, we check whether in the resulting (partial) world the consequent is true. We adopt here the particular approach proposed in [1]. According to this approach (and contra to [4], [5]) interventions are only possible for chancy events. Going back to the problem of fake Perfect in PP-Cs: I propose that the Perfect is used to locate the intervention in the past, while in SP-Cs the intervention is located at the present or in the future. I follow [6] in proposing that these restrictions on the location of the intervention are presuppositional. This idea can be also spelled out in terms of the modal approach to conditionals, but it uses a different modal base and ordering source than the standard approach based on the historical accessibility relation. This shift has a couple of advantages. Among other things, the approach can account for the central asymmetries observed in [3] between SP-Cs and PP-Cs (time asymmetry and presupposition asymmetry) without distinguishing two different evaluation times for conditionals. But most importantly, it allows us to account for the Morgenbesser cases like example (2), that are known to be problematic for PaP approaches. (2) A coin is going to be thrown and you have bet $5 on heads. Fortunately heads comes up and you win. You say: If I had bet on tails I would have lost. In an interventionist approach the outcome of the coin flip stays constant, because the event is causally independent from the antecedent. The conditional comes out as true. To sum up, combining a PaM approach to the Simple Past and a PaP approach to the Perfect in PP-Cs allows us to account for the distribution of SP-Cs and PP-Cs in case of generic counterfactuals. The switch to an interventionist perspective on the semantics of conditionals helps us to improve on shortcomings of other PaP approaches. Finally, this approach is able to account for the necessary counterfactuality of sentences like (1.c). As stated above, the interventionist approach can be translated into a standard restrictor approach. In the relevant ordering source interventions take place a close as possible to the antecedent. If we combine this with the presupposition carried by the ”fake” Perfect, we get that the closest possible intervention that made the antecedent true has been a past intervention. From this it follows that the antecedent has to be counterfactual. 2 References: [1] [2] [3] [4] [5] [6] Hiddleston (2005). A Causal Theory of Counterfactuals. Nous 39(4). Iatridou (2000). The grammatical ingredients of counterfactuality, Linguistic Inquiry, 31. Ippolito (2013). Subjunctive conditionals: a linguistic analysis. MiT Press. Pearl (2000). Causality. Cambridge University Press. Pearl (2013). Structural Counterfactuals: A Brief Introduction. Cognitive Science, 37. Schulz (2014). Fake Tense in conditional sentences. Natural Language Semantics, 22. Appendix: Empirical results Study 1 The first study is a corpus study using 10 articles published in the last 20 years on the subject of counterpossible conditionals.3 The corpus was annotated manually for form: ‘sp’ for SP-Cs, ‘pp’ for PP-Cs, ‘ind’ for indicative conditionals and ‘mix’ for cases in which antecedent and consequent differ in form. The table below gives the results for those conditionals in the corpus that did not refer to the past. sp counterfactuals percentages counterpossibles percentages total 175 147 sp 161 92% 133 90 % pp 4 2% 5 3% ind 5 3% 5 3% pp ind mixed cases 3% 3% 2% mix 5 3% 4 3% 92% Figure 1: Results of the 1st study, the diagram only displays the results for the first row: counterfactuals Study 2 The second study is an online study on generic counterfactuals (designed with Qualtrics, published on prolific.ac). Participants where asked to judge the naturalness of the SP-C form compared to the PP-C form. The questionnaire consisted of 12 sample sentences plus 24 fillers, distracting the participants from the conditional form and the generic content of the target sentences. The 52 participants could choose between the following answers: SP-C is the more natural form (SP-C), PP-C is the more natural form (PP-C), both forms are equally natural (both), none of the two forms sounds natural for the given example (none). 40 30 percentages SP-C 66 % PP-C 9% both 19 % none 7% 20 10 0 SP PP both none Figure 2: Results of the 2nd study 3 Counterpossible conditionals are conditionals with an antecedent that cannot possibly be true. A typical example would be ”If Hobbes had succeeded in squaring the circle, ...”. 3
© Copyright 2026 Paperzz