Relative past and the syntax - semantics interface in Russian

In this paper, I argue both empirically and formally that we need an optional relative past in the semantics of the temporal system of Russian. I propose a temporal calculus which can effectively deal with relative past readings in Russian, including the interaction with temporal adverbials, without positing an unwelcome ambiguity of explicit markers of tense and aspect. Furthermore, I show that a relative past is, to a certain extent, independent of aspect and cooccurs also with imperfective aspect, pace Paslawska and von Stechow (2003). More specifically, the so-called factual Ipf clearly exhibits past perfect interpretations, although it does not freely allow interpolation of a relative past. 1 Tense and aspect in a language without perfect tenses Russian lacks an overt expression for the kind of relative past expressed through the composite perfect in Germanic and Romance. From the point of view of morphology, tense proper is a simple story in Russian which basically can be subsumed as ‘±past’. However, the temporal system is significantly complicated by viewpoint aspect, that is, the morphological and semantic distinction between perfective (Pf) and imperfective (Ipf) verb forms. In this paper, I will focus on complete event interpretations of both Pf and Ipf (the notorious konstatacija fakta, “factual Ipf”). This fine-grained aspectual distinction in Slavic allows for an overt disambiguation of the ambiguous present perfect in English; cf. the translations below, where one and the same English form corresponds to the morphological realisation of two different grams in Russian: (1) Ja uže skosil_pf / kosil_ipf travu. I already mowe-PAST-PF / mowe-PAST-IPF lawn I have already mowed the lawn. (resultative perfect Pf; experiential perfect Ipf) The difference in aspectual choice in (1) is at the heart of aspectual competition in Russian. How should we capture the temporal configurations of these Russian sentences at the syntaxsemantics interface, given the standard assumption that viewpoint aspects encode temporal relations between the assertion time and the event time (Klein 1995)? Since Russian overtly expresses both tense and aspect, one would like to see, from the point of view of compositionality, a principled and systematic account of their interaction. 1.1 Against Borik’s ‘present perfect’-analysis of factual Ipf 1 I am grateful to Kjell Johan Sæbø and Arnim von Stechow for stimulating discussions on the formalisation of the phenomena discussed in this paper. Without the inspiring works of these two scholars, it would not have been possible to develop the theory outlined here. 2 Crucial for the understanding of (1) is the competition between the unmarked imperfective and Pf with its temporal anchoring (here: temporal anchoring to the result state holding at the utterance time). I refer the reader to (Grønn 2004) for an analysis. Concerning examples like (1) above, I will argue that the present perfect which shows up in the English translations needs not be captured by our formalisation of the Russian data. This follows from the fact that the present perfect and the simple past are truth-conditionally equivalent (with telic predicates) in languages like English. The perfect expresses a precedence relation between two reference times. Present perfect is characterised by the presence of certain consequences of the event holding at the utterance time s*, but the indexical s* is always available for the temporal interpretation, hence we do not need a present perfect to give us access to this parameter. Therefore, I will argue against the analysis of (Borik 2002), where the factual Ipf in constructions like (1) is dubbed “the present perfect of the Russian imperfective”. The morphology is a simple past and the interpretation should be a simple past. To make this point clear it will still be useful to have a closer look at Borik’s proposal, which has the merits of being explicitly stated and can therefore easily be tested. Borik presents a privative analysis of the aspectual system in Russian. The marked Pf is identified with two conditions which both must hold for a perfective sentence to be true, while Ipf is defined in negative terms as non-perfective. Thus, Ipf is used as the unmarked member if at least one of the two conditions on perfectivity is not met (Borik 2002, 8). The definition of Pf proposed by Borik is spelled out in (2) below (cf. Borik 2002, 193), where the variables have the classic Reichenbachian interpretation, except for the fact that they obviously refer to intervals rather than points in time: (2) (i) Pf in Russian is defined by the configuration: S R = Ø & INCLUDES(R, E) (ii) Ipf is defined as non-perfective, i.e.: ¬ [S R = Ø & INCLUDES(R, E)], hence S R Ø or ¬INCLUDES(R, E) The first condition on Pf, which says that the reference time should not overlap the utterance time, is primarily supposed to reflect the fact that Pf with present tense morphology has a future tense interpretation. Concerning aspect proper, note that the complete event interpretation, which amounts to the event time being included in the reference time, is the default interpretation in this system. Hence, if the tense-like condition (the relation between S and R) on perfectivity is violated, Ipf is automatically used with the possibility of leaving the aspectual configuration INCLUDES(R, E) unchanged. Of course, from a purely logical point of view, the system does not exclude the possibility of Ipf being triggered by a negation of both conjuncts in the definition of Pf. However, Borik seems to defend the view that INCLUDES(R, E) is true of Ipf by default when the first conjunct of the condition on perfectivity is negated. In a sense, this makes the progressive reading semantically marked compared to factual Ipf. Since factual Ipf denotes complete events, factual Ipf can, of course, only be defined negatively with respect to the first condition on Pf. Hence, according to Borik’s theory, the reference time and utterance time must overlap in the case of factual Ipf, that is, S R Ø. This is also why Borik refers to factual Ipf as the “Present Perfect reading of Ipf”. One problem with this analysis is the use of the reference time parameter R, which carries too many functions. In section 2, I propose to split up the notion of reference time into what I will call the assertion time, the evaluation time and the frame time. More importantly, concerning factual Ipf, Borik’s theory is problematic in view of the data presented in section 3, where it is empirically shown that factual Ipf cooccurs with past perfect readings. Hence, the utterance time is not the only possible evaluation time for a past tense in Russian, irrespective of viewpoint aspect. 1.2 The importance of temporal adverbials On closer inspection, the Russian constructions in (1) differ from their English translations not only with respect to the morphological make up. The analogy between resultative/experiential perfect and perfective/imperfective aspect soon breaks down. Consider for instance the alleged “experiential perfect” reading of the factual Ipf in (1). If we extend the range of data, several differences emerge. For instance, unlike the experiential perfect in English, factual Ipf in Russian is perfectly fine with frame adverbials: (3) Ja padal s dereva v detstve. I fall-PAST-IPF from tree in childhood I fell (# have fallen) from a tree in my childhood. Of course, there is nothing extraordinary about the Russian data, since the tense form being used is a simple past, and not a composite perfect. Factual Ipf in Russian (past tense + a complete event interpretation), unlike the prototypical experiential perfect in Germanic, therefore cooccurs happily with (sufficiently large) frame adverbials. Accounting for this interaction of frame adverbials and factual Ipf is straightforward, as soon as we leave behind the illusion of a perfect tense. In fact, frame adverbials are particularly important for any aspectual theory couched in temporal terms, since they are crucial in determining the value of reference times. In (3) above, the assertion time equals the frame time v detstve ‘in my childhood’ and makes the past tense redundant. In other cases, the assertion time is constrained by both a frame adverbial and past tense: (4) Segodnja ty obedal v restorane! (internet) Today you eat-PAST-IPF in restaurant Today you had dinner in a restaurant! The frame adverbial segodnja ‘today’ denotes an interval of 24 (or, perhaps, 12) hours, but the assertion time, which the speaker focuses on, is only the part of today which is located in the past, i.e. prior to the utterance time s*. In the next section, I will spell out more precisely how aspect, tense and temporal adverbials interact in Russian. 1.3 Some ingredients for a temporal calculus I assume a standard semantics for aspectual operators (e.g. Kratzer 1998), here spelled out as a DRS (Discourse Representation Structure) in the framework of -DRT (a compositional version of Discourse Representation Theory): (5) Pf (or factual Ipf) P t [e | P(e), INCLUDES(t,e)] By using -expressions to represent functions of different complexity, we can build the DRSs compositionally, working bottom-up. According to the formula in (5), viewpoint aspects map properties of events into properties of times. I thus assume that the aspectual operator combines through functional application with the aspectand tenseless VP. The VP, including internal and external nominal arguments, denotes a property of events in the spirit of neoDavidsonian semantics. Note also that aspectual operators are responsible for declaring the event variable in the universe of the DRS ( existential quantification). Building existential quantification of the event into the semantics of the aspectual operator is standard in current approaches to temporality at the syntax-semantics interface. Since grammatical aspect is obligatory in Russian, this implies that we do not need to invoke default existential closure of the event argument. In other words, the event argument in Russian is overtly realised in the morpho-syntax by the aspectual operator. Concerning semantics proper, aspectual operators establish purely temporal relations between the event time and the assertion time. Following Klein (1995, 669), the event time is the time at which some situation obtains, while the assertion time is the time which the speaker focuses on, i.e. “the time for which an assertion is made by the utterance which describes the situation”. The semantic contribution of Pf and factual Ipf is to assert that the assertion time t includes the time of the event e denoted by the predicate P, or, alternatively, e is included in t. In the present framework, this takes the form of a DRS condition making use of the two-place predicate INCLUDES. The progressive/processual reading of Ipf, on the contrary, can reasonably be seen as an instance of the opposite inclusion relation, which can be represented by switching the arguments of the topological relation: INCLUDES(e,t). The assertion time, i.e. the speaker’s focus, is necessary to make visible the situation talked about in a sentence, but where does it come from? The standard answer is that a tense operator having scope over aspect locates the assertion time in the past, present or future. However, as we saw above in section 1.2, temporal adverbials – particularly frame adverbials – play an important role in establishing the value of the assertion time. To capture this interaction and account for the intuition that tenses and adverbs together determine the value of the assertion time, I adopt a syncategorematic treatment of tense, known as the “definite theory of tense” (von Stechow 1995): (6) Past*(t)(s*) denotes the intersection between t and the past of s*, in other words, the maximal interval in t prior to s*. Tense itself is a function from a frame time t to an assertion time t’. This amounts to some kind of semantic unification, where tenses and temporal adverbials are part of the same ‘branch’ in the tree, a tree which ends in the assertion time interval at the top node. This provides us with a compositional account, through the assertion time parameter, of the interaction between aspect and other temporal phenomena. The final DRS for example (4) would thus be: (4’) [e | dinner(e), Agent(e, hearer), Place(e, restaurant), INCLUDES(Past*(today)(s*),e)] The representation in (4’) says that the event of having dinner is located in the interval resulting from intersecting ‘today’ and ‘all time preceding the utterance time’. Uttered at, say, 4 o'clock on July 1, the sentence would mean that the dinner took place some time on July 1 before 4 o'clock. In other words, the assertion time is only the part of today which is located in the past. This example therefore clearly shows that the assertion time cannot automatically be identified with an overt frame adverbial or with tense proper (past tense “the whole past”). 2 Relative past and perfective aspect In the previous section, I argued against distinguishing between so-called ‘present perfect’ and ‘simple past’ interpretations of past tense in Russian. The difference shows up in translations into English, but these cases should be modelled uniformly, reflecting the uniform tense morphology in Russian. Now comes the tricky part. What happens if the interval denoted by tense does not relate to the speech time s*? It is well-known that Pf and simple past morphology can occur with past perfect interpretations in Russian: (7) Kogda Karpov stal empionom mira, on uže vyigral empionat mira sredi junošej. When Karpov become-PAST-PF champion of-world, he already win-PAST-PF championship of-world among juniors. When Karpov became world-champion, he had already won the junior worldchampionship. Indeed, such examples represent a challenge to the uniform one-to-one correspondence between morpho-syntax and semantics, which was suggested above. The perfective verb vyigral ‘won’ denotes a complete event located in the past of the temporal adverbial clause, which in turn denotes an interval prior to the utterance time. The fact that past tense sentences in Russian cannot be reduced to a deictic past must affect our formalisation of temporality at the syntax-semantics interface. This argument is perhaps even more striking for examples with an overt future tense and a covert past, like the following, discussed in the works of Paslawska and von Stechow: (8) V sem’ asov Maša uže vyjdet iz doma. At seven hours Maša already leave-FUT-PF from house At seven, Maša will have left already. In example (8), the future tense relates the temporal adverbial v sem’ asov ‘at seven o’clock’ to the future of the utterance time, while a covert relative past seems necessary to account for the available reading whereby the event occurs before 7 o’clock. 2.1 Against Paslawska and von Stechow’s ambiguity of Pf The first authors to take seriously the formalisation at the syntax-semantics interface of relative past readings in Russian were Paslawska and von Stechow in partly unpublished works from 1999 and onwards. Their approach, however, is heavily biased towards Pf, which is claimed to be ambiguous between the usual complete event interpretation (the event is included in the assertion time) and a precedence relation (the event precedes the assertion time, which in turn precedes or follows the utterance time, depending on tense). The latter temporal relation indeed results in a relative past interpretation. The two available readings for (8) would be: (8’) [e | leave(e), Agent(e, Maša), INCLUDES(Fut*(7 o’clock)(s*),e)] An event of Maša’s leaving is included in the maximal future of 7 o’clock with respect to the utterance time, i.e. Maša’s leaving is included in the interval of 7 o’clock. (8’’) [e | leave(e), Agent(e, Maša), PRECEDES(e, Fut*(7 o’clock)(s*))] An event of Maša’s leaving precedes the maximal future of 7 o’clock with respect to the utterance time, i.e. Maša’s leaving precedes the interval of 7 o’clock. Paslawska and von Stechow’s account maintains a uniform treatment of the overt tense morphology, since the value of the “assertion time”, i.e. Fut*(7 o’clock)(s*), is the same in both cases. However, the ambiguous perfective aspect induces two different temporal configurations: the purely aspectual relation INCLUDES and the more tense-like relation PRECEDES. Similarly, in the case of past perfect readings of Pf, such as (7) above, Pf contributes the condition PRECEDES(e,t), while the overt past tense of vyigral ‘won’ serves to locate the temporal ‘kogda/when-clause’ in the past of the utterance time: Past*(when Karpov became world-champion)(s*). By combining the contribution of perfective aspect and past tense, we arrive at the correct past perfect configuration: PRECEDES(e, Past*(when Karpov became world-champion)(s*)). This system is compositional only if the relative past is consistently expressed by overt perfective morphology. However, the idea of linking the phenomenon of relative past to viewpoint operators (in particular Pf) runs into principled problems related to the interaction of tenses and temporal adverbials. One problem with the above analysis is that there is no room for temporal adverbials being true frame times of the event on a past perfect reading. This is also pointed out by the authors themselves: “temporal adverbs should modify times directly. So there might be something wrong with our system” (Paslawska and von Stechow 2003, 340). 2.2 The evaluation time parameter In a system where a relative past, represented as the relation PRECEDES(e,t), replaces the aspectual inclusion relations, we should no longer consider t to be the assertion time. It is well-known that Reichenbach’s original use of his ingenious ‘reference time’ is overloaded as it assumes too many functions at once. In this paper, I therefore refer to the reference time only informally as a cover term. In order to account for cases like (7) and (8), we need an additional reference time parameter, different from the assertion time, the frame time and the utterance time. The reference time in question is sometimes referred to in the literature as the evaluation time. The evaluation time refers to the temporal perspective of the speaker (interpreter). The default value of the evaluation time is the utterance time, but in (7) and (8) above the evaluation time is importantly different from both Reichenbach’s S and R. Following (von Stechow 1995), I therefore replace s* in the definition of past tense in (6) by the distinguished variable t0. The convention thus adopted is that t0 denotes the local evaluation time, which equals the utterance time whenever t0 is free. Henceforth, tense is the function in (9), a function from a frame time to an assertion time: