406 8. Temporal domains and relative tenses: theoretical foundations he would wait until his wife {arrived /*would arrive}. If we assume that the meaning of until includes an ‘implicit orientation time’ Ϫ until means ‘until the time that’ Ϫ the use of the past tense is explained naturally: it represents the situation time of John’s wife’s arrival as T-simultaneous with the implicit orientation time. 8.28.1 The concept of relative past tense is also necessary to account for the obligatory use of the preterite in the adverbial time clauses of examples like the following: Maud knew that John would leave before Mary {arrived /*would arrive}. As noted in 14.4.2, the before-clause in this example is interpreted as ‘not-yet- factual at the binding time’, more specifically as not-yet-factual at the situation time of the head clause: what Maud knew was that Mary would not yet have arrived when John left. The same applies to the until-clause in the following ex- ample: Maud expected that John would wait until Mary {arrived / *would arrive} [but she couldn’t know that John would be waiting in vain because Mary wouldn’t make it to the meeting-place that day.]. (The until-clause has a similar ‘not-yet-factual at the head clause-situation time’ interpretation) The obligatory use of the past tense in such time clauses is generally looked upon as puzzling. It seems illogical that (unlike many other languages) English does not allow the use of the conditional tense (which represents a situation time as T-posterior to an orientation time in a past domain) to refer to a situation time which must actually be interpreted as W-posterior to an orienta- tion time in a past domain. However, these data do not really present a prob- lem if one knows that the temporal conjunctions before, until, after, since and when resemble the phrasal conjunction by the time that in that they are inter- preted as ‘{before / until / after / since / at} the time that’ (see 8.2.1). Diachron- ically, the present-day conjunctions before, until, after, and since have actually developed from prepositional phrases of this kind. The fact that we must use arrived rather than would arrive in the above examples shows that, in an inten- sional context, the situation time of a before / until-clause has to be represented as T-simultaneous with the orientation time that is implicit in the semantics of the conjunction (and which is explicit in the conjunction by the time that). A comparison of the following sentences renders it easier to see this constraint: John left before Mary arrived. John decided he would wait until his wife {arrived / *would arrive}. The first example is interpreted as meaning that both situations actualized in the order indicated by before. Because of the factuality of the two situations, III. Arguments for distinguishing between the absolute and the relative past tense 407 the two preterites are analysed as absolute preterites in our theory. (As ex- plained in 14.5.1, not all before-clauses use a relative tense. Those that repre- sent a situation as a past fact use the absolute past tense.) The implication that Mary had not yet arrived when John left is an entailment but does not alter the fact that the two situations are represented as factual. In the second example, decided creates an intensional world in which the situation of John’s wife’s arrival is expected to actualize. The form arrived cannot be an absolute past tense here because the speaker does not represent the arrival as a past fact: it is quite possible that John’s wife did not actually show up. In other words, by using an absolute past tense a speaker always represents a situation as factual from his point of view, i. e. from the temporal standpoint of t 0 . Such a reading, which is given to both past tense forms in John left before Mary arrived, is a ‘transparent’ (‘de re’) interpretation. By contrast, in John decided he would wait until his wife arrived, arrived receives an ‘opaque’ (‘de dicto’) reading. The arrival only forms part of John’s inten- sional world, not of the speaker’s transparent world. In this world, the arrival is not-yet-factual at the time when John decided to wait. In sum, arrived is now a relative tense form receiving a ‘not-yet-factual at the binding time’ inter- pretation but no factual interpretation. (Unlike English, Germanic languages like Dutch or German use the conditional tense to express this not-yet-factual interpretation. In archaic English, until his wife should arrive could be used similarly instead of until his wife arrived. The following is an authentic exam- ple from the Bible: His lord was angry, and delivered him to the tormentors, until he should pay all that was due to him. 8.28.2 What we learn from the impossibility of using would arrive in the second example [John decided to wait until his wife arrived] is that the situa- tion time of an until-clause cannot be bound indirectly (by means of the use of the conditional tense) by the situation time which also binds the situation time of the head clause of the until-clause: the situation time of the arrival cannot be represented as T-posterior to the situation time of decided. The only pos- sibility is for the situation time of the until-clause to be bound directly by the orientation time which is implicit in the meaning of until (and which is lexi- calized as the time in the paraphrase ‘until the time that’). Since the relation in question is T-simultaneity Ϫ ‘until the time that his wife arrived’ means ‘until the time of his wife’s arrival’ Ϫ a relative past tense form must be used for this. (Remember that the relative past tense is the only tense that can be used to express T-simultaneity with an orientation time in a past domain.) That the past tense in question cannot be an absolute past tense is furthermore clear from the fact that an absolute preterite can only express one T-relation, viz. T- 408 8. Temporal domains and relative tenses: theoretical foundations anteriority to t 0 , 19 and from the fact that the orientation time that is implicit in until, as well as the situation time that is represented as T-simultaneous with it, can actually be W-posterior to t 0 : Yesterday evening John decided he would wait in the hotel until his wife arrived today. [She has not arrived yet, but she will in a couple of hours.] 8.28.3 It is worth noting that the principle prohibiting the use of would arrive in John decided he would wait until his wife {arrived /*would arrive} also underlies the obligatory use of the present tense (as ‘pseudo-absolute’ tense expressing T-simultaneity Ϫ see 9.17) in the post-present counterparts of these examples and in similar sentences with a before-clause: John has decided he will wait until his wife {arrives /*will arrive}. John will wait until his wife {arrives /*will arrive}. John will leave before Mary {arrives /*will arrive}. The present tense is used because before / until mean ‘before / until the time that’ and the tense of the time clauses has to express T-simultaneity with this implicit orientation time. Using the absolute tense form will arrive is not al- lowed: an adverbial before / until-clause cannot establish a post-present do- main of its own. There is a clear parallelism between the ungrammaticality of will arrive here and the ungrammaticality of would arrive in John decided he would wait until his wife {arrived /*would arrive}. In both cases the situation time of the time clause can only be T-related to the orientation time which is implicit in the meaning of the conjunction. Using will arrive or would arrive would T-relate it to another orientation time. (In the case of will arrive the situation time would be T-related directly to t 0 .) 8.29 Argument 7: ambiguity or vagueness? John said that Mary was ill allows a reading in which Mary was ill at the time of John’s utterance and one in which Mary was ill at some time preceding John’s utterance. Our claim that these two interpretations are a matter of ambiguity (tense structure) and not vagueness (W-interpretation) is confirmed by Lakoff’s well-known test to distinguish between ambiguity and vagueness. 19. As in all the arguments given here in support of the notion of relative past tense, we are disregarding special uses of past tense forms, as in He was leaving the next day, If you did that tomorrow …, If only I was a girl!, etc. By ‘absolute past tense form’ we always mean a tense form that locates a situation time in the past time-sphere and in doing so represents it as T-anterior to t 0 . III. Arguments for distinguishing between the absolute and the relative past tense 409 8.29.1 The above distinction between T-simultaneity and W-simultaneity is a distinction between semantic meaning and W-interpretation. (The former is exclusively a question of semantics, whereas the latter is a question of pragmat- ics as well as semantics: the interpretation of a clause depends not only on the semantics of the construction used and the lexical meaning of the constituents but also on pragmatic factors, such as the (linguistic and extralinguistic) context in which the clause is used, our pragmatic knowledge of the world, and the generally accepted principles of conversation.) The distinction between T-simultaneity and W-simultaneity is therefore related to the distinction be- tween (semantic) ambiguity and (pragmatic) vagueness. A sentence like John said that Mary was ill clearly allows two W-interpretations: (a) Mary was ill at the time of John’s utterance, and (b) Mary was ill at some time preceding John’s utterance. (Out of context, interpretation (a) is the normal one; reading (b) can be brought out, for example, by the addition of a time adverbial, as in [John said that] Mary was ill [when the plane landed].) Now, there are only two theoretical ways of explaining the fact that there can be two W-inter- pretations. The first is to say that on either reading was is an absolute past tense form, i. e. that like said, was expresses no more than that its situation time lies in the past time-sphere. The two readings of the sentence then follow from the fact that the two situation times may or may not be interpreted as W-simultaneous with each other: the intended reading will be clear from the context. This explanation is based on the notion of pragmatic vagueness, not semantic ambiguity. The second explanation relies on the claim that was is semantically ambiguous between an absolute past tense form and a relative past tense form. Because of the different semantics of these two tenses Ϫ they express different temporal structures Ϫ this explanation is based on the notion of semantic ambiguity, not pragmatic vagueness. 8.29.2 We can show that the semantic explanation is the correct one by using a well-known test for ambiguity. The rationale of this test is as follows. Take a sentence which yields two interpretations. Add a clause expressing similarity and involving a verbal pro-form, for example and so {did / was} Bill.Ifthe two interpretations of the initial sentence are a matter of ambiguity, the clause with the pro-form will have to receive the same interpretation as the sentence to which it is added. If this is not the case, i. e. if the initial sentence and the added clause allow a different interpretation, there is vagueness but not ambi- guity. For example: John likes Bill more than Mary, and so does Peter. The sentence John likes Bill more than Mary is syntactically (and therefore also semantically) ambiguous between two meanings: John likes Bill more than he likes Mary. John likes Bill more than Mary likes Bill. 410 8. Temporal domains and relative tenses: theoretical foundations As predicted, the added clause in John likes Bill more than Mary, and so does Peter must receive the same reading as the first clause. That is, of the following four readings, only the first two are possible interpretations of the sentence: John likes Bill more than he likes Mary, and Peter also likes Bill more than he likes Mary. John likes Bill more than Mary likes Bill, and Peter also likes Bill more than Mary likes Bill. John likes Bill more than he likes Mary, and Peter also likes Bill more than Mary likes Bill. John likes Bill more than Mary likes Bill, and Peter also likes Bill more than he likes Mary. When applying this test to John said that Mary was ill, we find that the double interpretation of this sentence is due to semantic ambiguity: John said Mary was ill, and so did Bill. Because John and Bill may or may not have spoken at the same time, the following are the possible interpretations of this sentence: John said [at t 1 ] that Mary was ill [at t 1 ], and Bill also said [at t 1 ] that Mary was ill [at t 1 ]. (In this case John and Bill spoke at the same time.) John said [at t 1 ] that Mary was ill [at t 1 ], and Bill said [at t 2 ] that Mary was ill [at t 2 ]’.) (In this case John and Bill spoke at different times.) John said Mary had been ill [at some earlier time], and Bill also said that Mary had been ill [at the earlier time John had in mind]. John said Mary had been ill [at some earlier time], and Bill also said that Mary had been ill [at some earlier time different from the earlier time which John had in mind]. (Because and so did Bill expresses similarity, this reading is less likely than the previous one.) The sentence John said Mary was ill, and so did Bill does not yield the follow- ing readings: *John said [at t 1 ] that Mary was ill [at t 1 ], and Bill also said [at t 1 or at t 2 ] that Mary had been ill [at some time earlier than either t 1 or t 2 ]. *John said [at t 1 ] that Mary had been ill [at some time earlier than t 1 ], and Bill also said [at t 1 ] that Mary was ill [at t 1 ]. *John said [at t 1 ] that Mary had been ill [at some time earlier than t 1 ], and Bill also said [at t 2 ] that Mary was ill [at t 2 ]. These sentences are actually ungrammatical because also requires that the two parts should have similar meanings, which they do not. The above data accord with the view that the two readings of Mary was ill in John said that Mary was ill result from semantic ambiguity: the form was III. Arguments for distinguishing between the absolute and the relative past tense 411 is ambiguous between two meanings. Readings resulting from vagueness are not subject to the restriction observed above. This is clear from the following: Bill was as ill as John was. In this example (which also expresses similarity) both verb forms are absolute past tense forms. Because both forms thus relate their situation time directly to t 0 , the temporal relation between the two situation times remains vague (unspecified): we do not know from Bill was as ill as John was whether Bill was ill while, before or after John was ill. It is therefore predictable that the test shows was to be vague rather than ambiguous between two temporal readings: Bill was as ill as John was, and so was Mary. Nothing whatsoever is implied here concerning the relative times of the three illnesses. Whichever of the three interpretations (viz. Bill was ill {while / be- fore / after} John was) we assign to the first part, all three interpretations (viz. ‘Mary was ill {while / before / after} John was’) remain possible in the second (ϭ added) part. 8.29.3 The test for semantic ambiguity vs vagueness that we have used here is due to Lakoff (1970). It has been argued in the literature that there are sentence types in which this test is not reliable. However the kind of sentence to which we have applied the test is not one of the problematic cases. The fact that the test works well in cases of indirect speech is further illustrated by the following: John will tell the chairman that Mary was too ill to attend the meeting. This sentence is semantically two-way ambiguous in that Mary’s illness may be taken to be T-anterior either to t 0 or to the post-present time of John’s speaking (i. e. the ‘pseudo-t 0 ’ Ϫ see 9.17.1). In the following sentences the reading to be selected is each time determined by the context: [Mary couldn’t attend yesterday’s meeting because she was ill. When I see him this afternoon] I will tell the chairman that Mary was too ill to attend the meeting. (Mary’s illness is anterior to t 0 . This means that was is an absolute preterite.) [If Mary becomes ill, she won’t be able to attend Saturday’s meeting. In that case I will speak to the chairman after the meeting.] I will tell him that Mary was too ill to attend the meeting. (Mary’s illness is anterior to the pseudo-t 0 but posterior to t 0 . This means that was is a ‘pseudo-absolute’ tense form (see 9.17.1Ϫ3) which actually functions as a relative tense expressing T-anteriority to the central situation time of the post-present domain.) If we build a sentence similar to John said Mary was ill, and so did Bill,we get the following: John will tell the chairman that Mary was too ill to attend the meeting, and so will Bill. 412 8. Temporal domains and relative tenses: theoretical foundations Of the four theoretically possible interpretations, only the first two are accept- able: John will tell the chairman that Mary was too ill to attend the meeting [at a time anterior to t 0 ], and Bill will also tell the chairman that Mary was too ill to attend the meeting [at a time anterior to t 0 ]. John will tell the chairman [at t 1 ] that Mary was too ill to attend the meeting [at a time anterior to t 1 but posterior to t 0 ], and Bill will also tell the chairman [at t 1 or at t 2 ] that Mary was too ill to attend the meeting [at a time anterior to t 1 or t 2 but posterior to t 0 ]. *John will tell the chairman [at t 1 ] that Mary was too ill to attend the meeting [at a time anterior to t 0 ], and Bill will also tell the chairman [at t 1 or at t 2 ] that Mary was too ill to attend the meeting [at a time anterior to t 1 or t 2 but posterior to t 0 ]. *John will tell the chairman [at t 1 ] that Mary was too ill to attend the meeting [at a time anterior to t 1 but posterior to t 0 ], and Bill will also tell the chairman that Mary was too ill to attend the meeting [at a time anterior to t 0 ]. Here too, the test works well. In this case it proves that there is a semantic difference between an absolute preterite (meaning ‘the situation time is past relative to t 0 ’) and a ‘pseudo-absolute’ preterite (meaning ‘the situation time is past relative to a post-present pseudo-t 0 ’). 8.30 Argument 8: choosing between the progressive and the nonprogressive form In John said that he was writing a book, the use of the progressive form is compulsory if we want to express the idea that at the time of his utterance John was engaged in the (telic) situation of writing a book. In the same way as we cannot say *I write a book, we cannot say *John said that he wrote a book (on the relevant interpretation). This can only be explained if we posit the existence of a relative past tense meaning ‘The situation time is T-simultaneous (ϭ strictly coinciding) with an orientation time in a past domain’. Since the time of saying (which functions as T-binding time) is (nearly) punctual, we need to represent the much longer situation of writing a book as homogeneous (by using the progressive), so that the full situation can be conceived of as consisting of representative subparts, one of which has the same restricted duration and temporal location as John’s utterance and can therefore function as the time of the predicated situation. 8.30.1 Barring special uses of the present tense (as in historical speech, sum- maries, etc., where there is a ‘shift of temporal perspective’ from the past to . clause of the until-clause: the situation time of the arrival cannot be represented as T-posterior to the situation time of decided. The only pos- sibility is for the situation time of the until-clause. is that the situa- tion time of an until-clause cannot be bound indirectly (by means of the use of the conditional tense) by the situation time which also binds the situation time of the head. However the kind of sentence to which we have applied the test is not one of the problematic cases. The fact that the test works well in cases of indirect speech is further illustrated by the following: John