Embedded tense interpretation and sequence of tense in Persian

Embedded tense interpretation and sequence of tense in Persian

Available online at www.sciencedirect.com ScienceDirect Lingua 226 (2019) 1--19 www.elsevier.com/locate/lingua Embedded tense interpretation and seq...

583KB Sizes 0 Downloads 89 Views

Available online at www.sciencedirect.com

ScienceDirect Lingua 226 (2019) 1--19 www.elsevier.com/locate/lingua

Embedded tense interpretation and sequence of tense in Persian Motahareh Sameri a, Gholamhossein Karimi-Doostan b,* b

a University of Tehran, Iran Department of Linguistics, Faculty of Letters and Humanities, University of Tehran, Tehran, Iran

Received 3 September 2018; received in revised form 21 May 2019; accepted 21 May 2019 Available online 12 June 2019

Abstract The present study aims at investigating tense interpretation in Persian complement and relative clauses and determining whether Persian is subsumed under the label of sequence of tense (henceforth SOT) or non-SOT languages. To achieve these aims, Persian past-under-past, present-under-past, and present-under-future constructions were examined. However, given that there was not a unanimous consensus between the researchers as to the interpretation of past-under-past constructions (which is one of the criteria for distinguishing SOT languages from non-SOT languages), 32 Persian native speakers were interviewed. Furthermore, since the nature of present tense and the SOT phenomenon are intertwined, an attempt was made to study the nature of Persian present tense. The in-depth analysis of the interview results and Persian data were indicative of the non-indexical nature of Persian present tense. Moreover, the results revealed that Persian, on account of exhibiting some SOT-like and some non-SOT-like behaviors, is a hybrid language defying the commonly-held SOT/non-SOT classification. Finally, using the de re theory, as proposed by Abusch (1997) and further developed by Ogihara and Sharvit (2012), a comprehensive analysis was provided for the simultaneous past reading of Persian past-under-past complement clauses. © 2019 Elsevier B.V. All rights reserved.

Keywords: SOT; Non-SOT; Persian present tense; Past-under-past; Indexicality

1. Introduction With regard to tense interpretation in embedded clauses, one complication which has attracted researchers’ attention is the interpretation of tense in constructions in which a past tense is embedded under another past tense (Enç, 1987; Abusch, 1997; Kusumoto, 1998; Ogihara, 1994, 1995a; Sharvit, 2003, 2014; Bary and Altshuler, 2014; Hatav, 2012; Stowell, 2007; Grønn and von Stechow, 2010, 2011, 2013; Zagona, 2013, among others). Researchers were particularly interested in past-under-past constructions because, in some languages, the embedded past tense is not a semantically interpretable past tense conveying anteriority; rather, it can be understood as a null, vacuous, or zero tense encoding simultaneity with the matrix clause past tense (dubbed the simultaneous past reading) (Sharvit, 2003, 2014; Zagona, 2013; Ogihara, 1995a; Hatav, 2012, etc.). This is illustrated by the following examples:

Abbreviations: ACC, accusative; PSP, past participle; ASP, aspect; SG, singular; NEG, negative; SUB, subjunctive; PROG, progressive. * Corresponding author. E-mail addresses: [email protected] (M. Sameri), [email protected] (G. Karimi-Doostan). https://doi.org/10.1016/j.lingua.2019.05.008 0024-3841/© 2019 Elsevier B.V. All rights reserved.

2

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

1) John said that Mary was pregnant. (ke) mæryæm dust-eš dašt. 2a) sale 2005 ʔæli goft Year 2005 Ali say-PAST (that) Maryam love-him have-PAST1 ‘In 2005, Ali said (that) Maryam loved him.’ (ke) mæryæm hæmun moq’e dust-eš dašt. 2b) sale 2005 ʔæli goft Year 2005 Ali say-PAST (that) Maryam that time love-him have-PAST ‘In 2005, Ali said (that) Maryam loved him then.’ (ke) mæryæm sale 2000 dust-eš dašt. 2c) sale 2005 ʔæli goft Year 2005 Ali say-PAST (that) Maryam year 2000 like-him have-PAST. ‘In 2005, Ali said (that) Maryam loved him in 2000.’ Example (1), being ambiguous, lends itself to two interpretations. The most salient interpretation of this sentence is the simultaneous past reading, in which the time of Mary's being pregnant is simultaneous with the time of John's saying. Besides this reading, it has a back-shifted reading, where the time of Mary's being pregnant precedes the time of John's saying. However, the back-shifted reading is not salient and is not attested by many speakers, if this sentence is not accompanied by a past-oriented adverb (Ogihara, 1995a; Boogaart, 1999). Example (2a), likewise reveals that, in Persian, past-under-past constructions can convey both the simultaneous past (2b) and the back-shifted readings (2c) which are brought to the fore by adding appropriate adverbs to the example. However, it is worth mentioning that sentence (2a) is not controversial and its interpretation is straightforward in that, as will be clarified in the following, there is not a unanimous consensus among Persian native speakers as to the interpretation of past-under-past constructions. As stated above, Persian and English past-under-past constructions can convey the simultaneous past reading. Nonetheless, not all languages exhibit the same behavior such that the Japanese (Ogihara, 1994, 1995a,b, 1996, 1999, etc.), Polish (Sharvit, 2014), and Hebrew2 (Hatav, 2010, 2012; Ogihara and Sharvit, 2012) counterparts of these sentences merely encode the back-shifted reading and the simultaneous past reading is encoded using present-underpast constructions. Languages such as English, whose past-under-past constructions express simultaneous past, are referred to as SOT languages. While, languages such as Japanese, whose past-under-past constructions in complement clauses do not encode the simultaneous past reading and encode back-shifted reading are dubbed non-SOT languages. The motivation behind this naming is that languages like English (whose past-under-past constructions convey the simultaneous past reading) are characterized by entailing an SOT rule which provides the simultaneous past reading encoded by these constructions. Different definitions have been put forward to determine what exactly the SOT rule is. For example, Ogihara (1995a, 1996) and Sharvit (2014, 2018) define it as a ‘‘tense deletion rule’’ which optionally deletes a tense that is locally ccommanded3 by a morphologically identical tense. Thus, it is argued that, in past-under-past constructions of SOT languages, the embedded past tense is deleted under agreement with its c-commanding tense and the simultaneous past reading is obtained. Abusch (1997) and Grønn and von Stechow (2010), on the other hand, consider the SOT rule as a ‘‘transmission rule’’ which transmits the temporal features of verbal quantifiers4 in SOT languages. For example, in English past-under-past constructions, the past tense of the matrix verbal quantifier is construed as being transmitted to the embedded clause tense. Thus, the tense of the embedded verb is determined by the matrix clause tense. Another important aspect in the discussion of past-under-past constructions is how the simultaneous past reading of these constructions is syntactically and semantically derived and accounted for. This issue is important because the embedded past tense, instead of conveying anteriority, conveys simultaneity (such as examples 1 and 2a). Sharvit (2018) argues that some approaches account for the simultaneous past reading of past-under-past constructions using the SOT rule; while, some other approaches such as the ‘‘de re’’ approach (as proposed by Abusch (1997)) and the ‘‘scalar theory’’

1

The list of all abbreviations is provided after the conclusion section. In Hebrew, a past tense embedded under another past tense in a complement clause, encodes the back-shifted reading for most Hebrew native speakers. However, for few Hebrew native speakers, such constructions encode the simultaneous past reading, too (Hatav, 2012; Ogihara and Sharvit, 2012). 3 According to a very broad definition, node A c-commands node B if A is not structurally higher than B and every node that dominates A dominates B too (Radford, 2006; Haegeman, 2008). 4 For Grønn and von Stechow (2010, 2012), verbal quantifiers refer to verbs abiding by the following conditions: being inflected with tense morphology and controlling an embedded tense's reference time. 2

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

3

(as proposed by Altshuler and Schwarzschild (2013a,b)) attempt to account for the simultaneous past reading of pastunder-past complement clauses (such as example 1) without using the SOT rule. Another issue which has attracted researcher's attention is accounting for the simultaneous past reading of pastunder-past constructions in some non-SOT languages, which, as their name indicates, are characterized by lacking the SOT rule. More precisely, it is argued that, in some traditionally-regarded non-SOT languages, such as Russian (Altshuler, 2004, 2008; Khomitsevich, 2007) and Hebrew (Hatav, 2012; Ogihara and Sharvit, 2012), some of past-underpast constructions encode the simultaneous past reading. Grønn and von Stechow (2010), Ogihara and Sharvit (2012), and Hatav (2012) (among others) provide persuasive analyses to account for the simultaneous past reading of pastunder-past constructions in such non-SOT languages, without appealing to the SOT rule. The present study has been motivated by a dearth of research on embedded tense construal in Persian, especially given the discussion above and the known cross-linguistic variation in embedded tense construal. Thus, this study sets out to investigate tense interpretation in Persian embedded clauses, determine whether Persian can be classified as an SOT or non-SOT language, and provide a comprehensive analysis for the simultaneous past reading of Persian pastunder-past constructions in complement clauses. This article is structured as follows: Section 2 provides a brief sketch of different parts of the interview and explicates the obtained results. Section 3 examines the nature of Persian present tense by comparing it with its English and Japanese counterparts. Section 4 is dedicated to investigating tense construal in Persian past-under-past, present-underpast, and present-under-future complement and relative clauses. Section 5 provides a cross-linguistic comparison of embedded tense interpretation in Persian, English, Japanese, and Hebrew. Section 6 examines whether Persian is categorized among SOT or non-SOT languages. In Section 7, an attempt is made to provide a persuasive account of the simultaneous past reading of Persian past-under-past constructions. Finally, Section 8 concludes the paper. 2. Data collection As mentioned above, the present study pursues scrutinizing tense interpretation in Persian complement and relative clauses and determining whether Persian should be categorized as an SOT or non-SOT language. However, the researchers do not concur with each other regarding the temporal interpretation of past-under-past complement clauses which proves to be instrumental in deciding whether Persian can be categorized as an SOT or non-SOT language. In fact, all the different criteria which have been proposed for distinguishing SOT languages from non-SOT languages can be reduced to whether a given language's past-under-past constructions allow the simultaneous past reading or not (Sharvit, 2003, 2014; Ogihara, 1995a; Hatav, 2012, etc.). Thus, to gain a better understanding of temporal construal in Persian past-under-past constructions and to discuss based on reliable data, 32 Persian native speakers were interviewed.5 Furthermore, the main goal of the interview was to gain some insights into the nature of Persian present tense and investigating tense shift in transforming direct speech into indirect speech. Participants of the present study were selected from different cities of Iran, using purposive sampling in that, to control the effect of native language, deliberate attempts were made to select interviewees with Persian as their mother tongue. Thus, Iranian Kurds, Turks, and Arabs were excluded from the present study. Half of the interviewees were male and half of them were female and their mean age was 28 years. In terms of education, 12 of the participants were Ph.D. students, 10 of them were M.A. students, and 10 of them were B.A. students. After developing the interview questions, we piloted them with 10 Persian native speakers to determine how the interview should be performed and whether any questions need to be revised, eliminated, or added. Then, based on the pilot study, the following revisions were made to the questions: the adverb ‘‘right then’’ was added to some of fill in the blank questions to highlight the simultaneous reading of the embedded and matrix clause tenses and context was provided for some of vague or ambiguous questions. Furthermore, interviewers gained some information regarding how the questions should be asked to address the issue the question really intended to address. The first part of the interview which included open-ended questions was dedicated to eight past-under-past constructions with saying (such as examples 3 and 4) and attitudinal (such as examples 5 and 6) verbs in the matrix clause. The interviewees were asked to judge whether these sentences encoded simultaneous past reading in Persian or not. It is worth noting that, the reason behind selecting verbs of saying and attitude is that, before carrying out the interview, the researchers hypothesized that past-under-past complement clauses with verbs of saying in the matrix clause do not allow simultaneous past reading; while, past-under-past complement clauses with attitudinal verbs convey the simultaneous past reading. Furthermore, because verbs of saying and attitude differ in terms of aspect (Zagona, 2004a,b,

5 It is worth noting that the results obtained from 32 participates are not sufficient to draw definitive conclusions about tense construal in Persian past-under-past constructions but they provide some insights into the meninges these Persian constructions can encode and allow us to make our arguments based on reliable data obtained from interviewing a number of Persian native speakers.

4

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

2005), they can reveal the role of the aspectual class of the matrix verb in the expression of simultaneous past reading of Persian past-under-past constructions. From the first part of the interview, the following four examples are provided for illustration. 3) ʔæli goft (ke) mæryæm bardar bud. Ali say-PAST (that) Maryam pregnant be-PAST. ‘Ali said that Maryam was pregnant.’ (ke) mæryæm xeili dust-eš dašt. 4) ʔæli goft Ali say-PAST (that) Maryam very love-him have-PAST ‘Ali said that Maryam loved him very much.’ 5) ʔæli bavær dašt (ke) mæryæm bardar bud. Ali belief have-PAST (that) Maryam pregnant be-PAST ‘Ali believed that Maryam was pregnant.’ 6) ʔæli mo’tæqed bud (ke) mæryæm xeili dust-eš dašt. Ali belief be-PAST (that) Maryam very love-him have-PAST ‘Ali believed that Maryam loved him very much.’ The second part of the interview was devoted to eight fill in the blank questions to examine whether present-under-past or past-under-past constructions are used for the expression of simultaneity. As mentioned above, in non-SOT languages, for the expression of simultaneity in past, a present-under-past construction is used (without necessarily requiring the embedded present tense to refer to both the matrix clause past tense and the utterance time (hereafter UT) or have the double access reading (henceforth DAR)6) and past-under-past constructions merely convey the back-shifted reading. While, in SOT languages, a past-under-past construction is used for the expression of simultaneous past. The reason for using two types of questions (open-ended and fill in the blank questions) is that viewing the same issue from different perspectives makes it more amenable to solution and provides a more comprehensive understanding of the issue under investigation. From the second part of the interview, the following examples are mentioned for further clarification: 7) ʔæli væqti danešju bud goft (ke) mæryæm hæmun moq’e dust-eš . . . Ali when student be-PAST say-PAST (that) Maryam that time love-him . . . ‘When Ali was a student, he said that Maryam . . .. . ... him then.’ (love) (ke) mæryæm hæmun moq’e ʔæfsordegi . . .. . ... 8) ʔæli qæble ʔezdevaj goft Ali before marriage say-PAST (that) Maryam that time depression . . .. . .. . .. ‘Before getting married, Ali said that Maryam . . .. . .. . . depression then.’ (have) væqti nojævan bud-æm bavær dašt-æm (ke) 9) mæn I when young be-PAST-1SG belief have-PAST-1SG (that) mohammæd mæn-ro dust . . .. . .. . . Mohammad I-ACC love . . .. . .. . . ‘When I was young, I believed that Mohammad . . .. . .. . .. . ... me.’ (love) (ke) mæryæm hæmun moq’e jasusi . . .. . .. . .. 10) dæh sal piš ʔæli fekr mi-kærd Ten year ago Ali think ASP-do-PAST (that) Maryam that time spy . . .. . .. . ... ‘Ten years ago, Ali thought that Maryam . . .. . .. . .. espionage then.’ (conduct) With regard to verbs of saying, the picture that unfolds for open-ended questions (e.g. examples 3 and 4) is as follows: most of the interviewees (23 participants (about 72%)) concurred with the view that past-under-past constructions are

6 According to a very rough definition, double access reading arises when a present tense is embedded under a past tense and the embedded clause present tense can hold throughout (i) matrix clause tense (or the time of attitude) and (ii) the actual UT. However, the provided explanation is very crude and some linguists such as Bary and Altshuler (2014) have made some modifications to the above definition.

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

5

Fig. 1. Percentage of simultaneous past reading of past-under-past complements in open ended questions.

either devoid of the simultaneous past reading or this reading is not the salient interpretation of these constructions. For example, according to most of the interviewees, the time of Maryam's pregnancy or Maryam's loving in examples (3) and (4), respectively, is not simultaneous with the time of saying in the matrix clause and, at the time of saying, Maryam no longer is pregnant or loves Ali. This might offer sufficient ground for arguing that past-under-past constructions with saying matrix verb do not allow the simultaneous past reading and; accordingly, claim that, in Persian, allowing or disallowing simultaneous past is a function of verb type (or the lexical aspect of the verb in the matrix clause). However, this claim is dubious in that, in fill-in-the-blank questions, to express simultaneity with a matrix clause saying verb (e.g. examples 7 and 8), 28 interviewees (87.5%) used past tense at least once to fill in the blanks. For example, the blank in examples (7) was filled in with past tense dust-eš dašt ‘loved’ more frequently than present tense dust-eš dar-æd ‘loves’. The same holds true for example (8), in which the frequency of the past tense ʔæfsordegi dašt ‘was depressed’ was more than the frequency of the present tense ʔæfsordegi daræd ‘is depressed’. This suggests that, in Persian complement clauses, both past-under-past and present-under-past constructions can be used for the expression of simultaneous past. As for verbs of attitude, in 23 (about 72%) of the interviews, to express simultaneity in past in fill in the blank questions, either the present tense is used in all of the blanks or it is used more frequently than the past tense. For instance, in example (9), present tense dust dare, ‘loves’ was used more frequently than the past tense dust dašt, ‘loved’ to fill in the blank. However, in open-ended questions (e.g. examples 5 and 6), 29 of the interviewees (90%) mentioned that pastunder-past constructions can be expressive of simultaneous past reading. This strengthens the argument that, in Persian complement clauses, both past-under-past and present-under-past constructions allow the simultaneous past reading. The results of the interviews are represented in the following figures. Fig. 1 depicts the results of the open-ended questions. As can be seen in Fig. 1, for saying verbs in the matrix clause, only 28% of the interviewees considered the simultaneous past reading for past-under-past constructions in the complement clause; while, for verbs of attitude, 90% of the interviewees considered a simultaneous past reading for thus-construction. Fig. 2 illustrates the percentage of participants who used present tense more frequently than past tense to express simultaneity with the matrix clause past tense in fill in the blank questions. As can be seen in Fig. 2, for verbs of saying, 44% used present tense more frequently than the past tense and, for verbs of attitude, 72% used present tense more frequently. Hence, it can be inferred that, for verbs of saying, 56% used past tense more frequently to fill in the blank and, for verbs of attitude, 28% used the past tense more frequently. In the third part of the interview, the interviewees were asked to transform a direct speech into an indirect speech to find out whether tense shift occurs in Persian or not. 11) Ali to Maryam: cera næ-yumæd-i konfærans? 5 dæqiqe dige šoru’ mi-š-e Why NEG-come-2SG conference? 5 minutes other start ASP-become-3SG ‘Why did not you come to the conference? It will start in 5 minutes.’

6

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

Fig. 2. Percentage of participants who used present tense more frequently in fill-in-the-blank questions.

Maryam to Ali (direct speech):

mæn I

mæriz sick

hæst-æm væ be-PRESENT-1SG and

ne-mi-y-am. NEG-ASP-come-PRESENT-1SG. ‘I am sick and I do not come.’ (The next day): Ahmad to Ali: cera mæryæm diruz næ-yum-æd konfærans? Why Maraym yesterday NEG-come-PAST-3SG conference? ‘Why did not Maryam come to the conference yesterday?’ Ali to Ahmad (indirect speech): beh-eš zæng zæd-æm goft . . . To-her call hit-PAST-1SG say-PAST. . .. . .. . .. . .. . ... ‘I called her and she said . . .. . .. . .. . .. . .’ The results of this section suggest that, in transforming direct speech into indirect speech, there is no tense shift in Persian in that none of the interviewees changed the embedded clause tense to agree with the matrix clause tense. In fact, in the above example, given that a present tense is used in the direct speech, in the indirect speech, likewise, a present tense is deployed. The verbs in the direct and indirect speech only differ in terms of their person such that the direct speech verb is conjugated in the first person singular; while, the indirect speech is conjugated in the third person singular. Finally, given that the SOT phenomenon and the nature of present tense are intertwined (Sharvit, 2003, 2014; Ogihara, 1995a), the last part of the interview, including two true/false questions and one fill in the blank question, was devoted to identifying the nature of Persian present tense in terms of matrix indexicality. Sharvit (2003: 672) defines matrix indexicality as follows: ‘‘in some languages, a present tense morpheme in present-under-past sentences must refer to the utterance time’’. Sharvit's (2003) definition of matrix indexicality is based on Schlenker(1999) definition and Schlenker (1999) conception of the indexicality of English present tense is inspired by Kaplan's (1977) contention that indexicals are obligatorily dependent on the actual context of the utterance. Hence, given that English present tense obligatorily refers to the actual UT (even in present-under-past constructions), Schlenker(1999) considers it as indexical. Enç (1987: 649), in the same line of thought, distinguishes languages based on whether their present tense denotes the UT or not and maintains that, in English, the present tense necessarily denotes the UT; while, Russian present tense does not denote the UT. This explicates why in English, the truth of present-under-past constructions depends on the simultaneity of the embedded present tense with the actual UT and conveying the DAR. However, in Russian present-under-past constructions, the embedded present tense is not required to express simultaneity with the actual UT and encodes simultaneity with the matrix clause past tense. The author refers to this characteristic of the present tense in languages

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

7

Diagram 1.

Diagram 2.

such as English as ‘indexicality’. To determine whether the present tense in Persian is indexical, we examined the responses from the following examples (12--14). 12) do sal piš ʔæli goft mæryæm bardar ʔæst (true or false). Two year ago Ali say-PAST Maryam pregnant be-PRESENT-3SG. ‘Two years ago, Ali said that Maryam is pregnant.’ 13) 2000 sal piš Yosef fæhmid Zuleika dust-eš dar-æd (true or false). 2000 year ago, Yosef find out-PAST Zuleika love-him have-PERSENT-3SG ‘2000 years ago, Yosef found out that Zuleika loves him.’ 14) mæryæm ʔælaln ʔæli-o dust næ-dar-æd ʔæma do sal piš ʔu-ra Maryam now Ali-ACC love NEG-have-3SG but 2 year ago he-ACC xeili dust dašt væ ʔæli fæhmid mæryæm hæmun moq’e . . . very love have-PAST and Ali find out-PAST Maryam that time love-him . . . ‘Maryam does not love Ali now but two years ago, she loved him very much and Ali found out that Maryam . . .. . . him then.’ (love) The results of this section revealed that 31 of the interviewees regarded the sentences in (12) and (13) as being true in Persian and used a present tense to fill in the blank in question (14). According to the definition of indexicality, we can claim that the answers of the above questions are suggestive of the non-indexicality of Persian present tense in that, in none of the questions in 12--14, the present tense is necessarily required to refer to the actual UT. To put it more precisely, the acceptability of example (12) is indicative of the non-indexical nature of Persian present tense. The reason is that if it were indexical and necessarily required simultaneity with the UT, the pregnancy would have to last two years to cover both the time of ‘‘saying’’ and the UT and this led to the oddity of the sentence in that a two-year pregnancy contradicts the natural state of affairs. Nonetheless, in English, with an indexical present tense, the equivalent of example (12) (i.e. example 15) is ungrammatical since the pregnancy ‘‘has to’’ include the time of saying and the actual UT and it is not consistent with the duration of pregnancy in the actual world (Sharvit, 2003; Bary and Altshuler, 2014; Zagona, 2004a,b; among others). 15) *Two years ago, John said that Mary is pregnant. In this respect, Zagona (2004a,b), arguing for the indexical nature of English present tense, maintains that English present-under-past constructions must encode DAR to be acceptable and an obligatory inclusion relationship should be established among the embedded present tense, the matrix past tense, and the UT. Thus, in accounting for the inaccuracy of example (15), she takes the indexical nature of English present tense for granted and attributes its oddity to the failure of the embedded present tense to take the matrix clause past tense into account (accordingly, violating the DAR requirements (Diagram 1)). However, in Persian, unlike in English, in the equivalent of this sentence (example 12), the present tense includes the embedded clause past tense and does not include the UT (Diagram 2). In line with the discussion maintained so far, the answers provided for the true/false question in (13), substantiate the non-indexicality of Persian present tense. The reason is that the acceptability of this sentence reveals that Zuleika's loving time is merely simultaneous with Yosef's finding out time and is not required to refer to the actual UT. If Persian present tense were indexical and necessitated simultaneity with UT, Zuleika's loving time would be required to cover both the

8

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

finding out time and the actual UT but because Zuleika, as a normal human being, does not live that long, the accuracy of this sentence would be doubted. The equivalent of this sentence in English (example 16) and languages such as English, which have indexical present tense, is unacceptable because, as mentioned above, Zuleika's loving is required to last for 2000 years (to cover both the UT and the time of finding out) which is not possible, given the normal human life span. 16) *2000 years ago, Yosef found out that Zuleika loves him. The fill-in-the-blank question in (14), likewise, confirms the non-indexical nature of Persian present tense in that, although, in the first part of the sentence, it is highlighted that Maryam does not love Ali at UT, a present tense is used in the most embedded part of the sentence to fill in the blank. Hence, in this sentence, the present tense does not refer to the actual UT and is used to express simultaneity with the immediately c-commanding tense, i.e. the time of ‘‘finding out’’. The English equivalent of this sentence in (17) is construed as being unacceptable in that although the first sentence rules out the possibility of simultaneity with the UT (which constitutes an indispensable part of English present tense meaning), a present tense is used in the most embedded clause. 17) * Mary does not love John now but she did two years ago, and John found out that Mary loves him.

3. The nature of Persian present tense In the previous section, we examined Persian present tense through interview questions. In the following, we will examine it further both by investigating its distribution in before-clauses and by comparing it with other languages. The distribution of tenses in before-clauses can provide some insights into the nature of present tense in different languages. In this regard, Sharvit (2014) argues that, in languages whose present tense non-indexical, present-underpast before-clauses are well-formed; while, in languages whose present tense is indexical, such constructions are ill-formed. Given that Persian allows present-under-past before-clauses (example 18) and disallows past-under-past constructions in before-clauses (example 19), we can conclude that Persian present tense is not indexical. The ungrammaticality of present-under-past in before-clauses of languages with indexical present tense can be explicated by the following English example. In the sentence ‘‘*I sang before John plays’’, the time of John's playing must overlap the UT and my past singing precedes the UT. This interpretation can be obtained from the meaning of tenses themselves; without taking the semantic contribution of ‘‘before’’ into account. 18) qæbl ʔæz inke ʔæli be-ræv-æd, mæn resid-æm.7 Before from that Ali SUBJ-go-PRESENT-1SG, I arrive-PAST-1SG. Intended meaning: ‘I arrived before Ali left.’ 19) *qæbl ʔæz inke ʔæli ræft/ræft-e baš-æd, mæn resid-æm. Before from that Ali go-PAST/go-PSP SUB-be-3SG, I arrive-PAST-1SG. ‘I arrived before Ali left.’ Based on the interview questions and tense distribution in before-clauses, we came to the conclusion that Persian present tense is non-indexical. However, a careful comparison of present tense in Persian, English, and Japanese reveals that, in some respects, Persian present tense is like its English and Japanese counterparts and, in some respects, it differs from them. The similarity between Persian and English present tenses is allowing the DAR in present-under-past complement clauses (example 20) and their difference is that, in Persian, unlike in English, the accuracy of present-underpast complement clauses is not dependent upon encoding the DAR (examples 12 and 21). The similarity between Persian and Japanese present tenses is being non-indexical and not requiring simultaneity with the UT in present-under-past complement clauses. Their difference is that, in present-under-past constructions, Persian present tense can convey both the DAR and simultaneity with the matrix clause past tense; while, Japanese present tense merely encodes simultaneity with the matrix clause past tense (Sharvit, 2014; Ogihara, 1995ab, 1999, etc.). The other difference between Persian and

7 it is worth noting that, in Persian before-clauses, a present tense, with subjunctive marking, is used. Persian present subjunctive is marked by the prefix ‘‘be’’ attached to the present stem of the verb. In examples 33 and 34, it is revealed that, in Persian before-clauses, present subjunctive can be used; while, past subjunctive (consisting of past participle of the main verb and the subjunctive form of ‘‘budæn’’ (to be), i.e. baš) cannot be used. As Sharvit (2014) argues, in Spanish, also a subjunctively-marked verb is used in the before-clause. Hence, in this regard, Persian is like Spanish. However, in Spanish, past-under-past is allowed in the before-clause.

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

9

Japanese is that, in Persian complement clauses, the simultaneous past reading can be expressed using both the present and past tenses; while, in Japanese, only a present tense can be used for this purpose. 20) ʔæli goft mæryæm bardar ʔæst. Ali say-PAST Maryam pregnant be-PRESENT. ‘Ali said that Maryam is pregnant.’ sal piš ʔæli goft lændæn zendegi mi-kon-e ʔæma ʔæmir diruz 21) do two year ago Ali say-PAST London live ASP-do-3SG but Amir yesterday goft dige ʔunja ni-st væ færanse zendegi mi-kon-e. say-PAST no longer there NEG-be-PRESENT and France live ASP-do-3SG. ‘Two years ago, Ali said that he lived in London but, yesterday, Amir said that he no longer lives there and now lives in France.’ Sentence (20) expresses the DAR in that the pregnancy can be conceived of as covering both the time of saying in the matrix clause and the UT. However, the acceptability of this sentence does not hinge on encoding the DAR and this is demonstrated by the accuracy of the sentence in (12) in which the pregnancy is merely simultaneous with the matrix clause tense. In example (21), likewise, although a present tense is used in the sentence immediately embedded below say-PAST, ‘‘living in London’’ is not necessarily required to overlap the actual UT and this is substantiated by the rest of the sentences highlighting that the mentioned state of affairs does not hold any longer (or at the UT). Finally, it is worth mentioning that the claim that Persian present tense is not indexical provides evidence in favor of subsuming Persian as a non-SOT language (refer to Section 6.3). Furthermore, the non-indexical nature of Persian present tense reveals one point regarding the grammatical structure of Persian before-clauses in that, as mentioned above, it explicates why in Persian before-clauses present tense is allowed; while, past tense is not. 4. The construal of tense in Persian complement and relative clauses Having reviewed the interview questions, their results, and the nature of Persian present tense, in this section, we will investigate tense construal in Persian past-under-past, present-under-past, and present-under-future complement and relative clauses. 4.1. The construal of past-under-past in Persian complement and relative clauses As mentioned in the previous section, in Persian complement clauses, past-under-past constructions (irrespective of whether the matrix clause is a verb of attitude or saying), can allow both the simultaneous past and back-shifted readings. This argument was explicated by examples (2b) and (2c). It was further substantiated not only by the openended questions (such as examples 3--6) which were construed as encoding the simultaneous past reading but also by the fill-in-the-blank questions (such as examples 7--10) in which past tense (in tandem with present tense) was used for the expression of simultaneity with the matrix clause past tense. For further clarification, consider the following example (22a) in which, as demonstrated by the adverbs, the embedded clause past tense can be either simultaneous with the matrix clause past tense (22b) or anterior to the matrix clause past tense (22c). Furthermore, this example also demonstrates that, in Persian past-under-past complement clauses, besides verbs of attitude and saying, factive verbs in the matrix clause convey the simultaneous past reading. Because there was no controversy over the simultaneous past reading of past-under-past complement clauses with factive verbs in the matrix clause, they were not included in the interview. 22a) to mi-dunest-i mæn ʔæz dæst-et narahæt bud-æm. You ASP-know-PAST-2SG I from hand-your sad was-1SG. ‘you knew that I was upset with you.’ diruz tu mehmuni mi-dunest-i mæn hæmun moq’e 22b) to You yesterday in party ASP-know-PAST-2SG I that time ʔæz dæst-et narahæt bud-æm. from hand-your sad was-1SG ‘Yesterday, in the party, you knew that I was upset with you then.’

10

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

22c) to diruz mi-dunest-i mæn hæfte piš ʔæz dæst-et narahæt bud-æm You yesterday ASP-know-PAST-2SG I week ago from hand-your sad was-1SG ‘Yesterday you knew that I was upset with you last week.’ Persian past-under-past relative clauses such as example (23a) convey the ‘‘simultaneous’’, ‘‘back-shifted’’, and ‘‘forward-shifted’’ readings. These interpretations are possible because of the adverbs. 23a) sale 1385 ʔæli zæni-ra did ke ʔu-ra dust dašt. Year 1385 Ali woman-ACC see-PAST that he-ACC love have-PAST. ‘In 2006, Ali saw a woman who loved him.’ did ke hæmun moq’e ʔu-ra dust dašt. 23b) sale 1385, ʔæli zæni-ra Year 1385, Ali woman-ACC see-PAST that that time he-ACC love have-PAST ‘In 2006, Ali saw a woman who loved him then.’ did ke sale 1360 ʔu-ra dust dašt. 23c) sale 1385, ʔæli zæni-ra Year 1385, Ali woman-ACC see-PAST that year 1360 he-ACC love have-PAST ‘In 2006, Ali saw a woman who loved him in 1981.’ did ke sale 1390 ʔu-ra dust dašt. 23d) sale 1385, ʔæli zæni-ra Year 1385, Ali woman-ACC see-PAST that year 1390 he-ACC love have-PAST ‘In 2006, Ali saw a woman who loved him in 2011.’

4.2. The construal of present-under-past in Persian complement and relative clauses Persian present-under-past complement clauses convey the simultaneous past reading because in the fill-in-the-blank questions of the interview, present tense (along with past tense) was deployed for the expression of simultaneity with matrix clause past tense. The other key observation regarding Persian present-under-past complement clauses is that, in these sentences (such as example 24), the embedded present tense need not refer to the actual UT. This was shown by the fact that a present tense was used by almost all of the interviewees (31 of them) in example (14). As mentioned earlier, in example (14), although it is clarified that Maryam does not love Ali at UT, a present tense is used in the most embedded clause. The most embedded present tense does not refer to the UT and encodes simultaneity with the immediately c-commanding past tense. 24) ʔæli fæhmid mæryæm dust-eš dar-æd. Ali find out-PAST Maryam love-him have-PRESENT-3SG. ‘Ali found out that Maryam loves him.’ Before proceeding, one issue should be clarified. Although, in example (14), 31 of the interviewees used a present tense to fill in the blank, a past tense can also be used in its most embedded clause. The observation that both present and past tenses can be used in the blank is corroborative of the claim that, in Persian complement clauses, both past-underpast and present-under-past constructions can be used for the expression of simultaneous past reading. Persian present-under-past relative clauses, as the following example demonstrates, do not encode simultaneity with the matrix clause past tense and merely express simultaneity with the actual UT. 25) ʔæli mærdi-ra did ke dar-æd gerye mi-kon-æd. Ali man-ACC see-PAST that be-PRESENT PROG cry ASP-do-1SG. ‘Ali saw the man who was crying.’ Example (25) does not encode a simultaneous past reading and merely expresses simultaneity with the UT. In fact, in this example, to express simultaneity with the matrix clause past tense, a past tense should be used. The observation that Persian present tense in present-under-past relative clause is always evaluated with respect to the actual UT is not indicative of the indexical nature of present tense in Persian. Because, in languages with proven non-indexical present tense (such as Japanese), the present tense in relative clauses can also refer to the actual UT (Ogihara and Sharvit, 2012: 644). Furthermore, tense shift (shifting the evaluation time from the default speech time to a different time such as an

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

11

attitude holder's now) is confined to intentional contexts and relative clauses are not among intentional contexts (Hatav, 2012; Abusch, 1997; Schlenker, 2003).

4.3. The construal of present-under-future in Persian complement and relative clauses In Persian present-under-future complement clauses (example 26), the embedded present tense is not required to express simultaneity with the actual UT. This argument is buttressed by the acceptability of the sentence in (27) in which it is mentioned, at the outset of the sentence, that Maryam does not love Ali at the time of UT. However, in the most embedded clause, to express simultaneity with the immediately c-commanding future tense, a present tense is deployed. This observation can be used as further evidence confirming the non-indexical nature of Persian present tense. 26) ʔæli xah-æd fæhmid ke mæryæm dust-eš dar-æd. Ali will-3SG find out-PAST that Maryam love-him have-3SG. ‘Ali will find out that Maryam loves him.’ 27) mæryæm ʔælan ʔæli-ro dust næ-dar-e ʔæma ʔæli inqædr xube ke mæryæm Maryam now Ali-ACC love NEG-have-3SG but Ali so good that Maryam do sal dige ʔašeq-eš mi-š-e væ ʔæli xah-æd fæhmid two year next love-him ASP-become-3SG and Ali will-3SG find-out-PAST ke mæryæm dust-eš dar-e. that Maryam love-him have-PRESENT. ‘Maryam does not love Ali now but Ali is so good that Maryam will love him in the next two years and Ali will find out that Maryam loves him.’ In Persian relative clauses with a present tense under a future tense, the embedded present tense need not refer to the actual UT and is merely required to express simultaneity with the matrix clause future tense, which is demonstrated by the following examples in (28) and (29). 28) ʔæli zæni-ra xah-æd did ke ʔu-ra dust dar-æd. Ali woman-ACC will-3SG see-PAST that he-ACC love have-PRESENT-3SG ‘Ali will see a woman who loves him.’ xah-æd did ke ʔælan ʔu-ra 29) ʔæli zæani-ra Ali woman-ACC will-3SG see-PAST that now he-ACC dust næ-dar-æd ʔæma dær ʔayænde ʔu-ra dust dar-æd. love NEG-have-3SG but in future he-ACC love have-PRESENT. ‘Ali will see a woman who does not love him now but will love him in the future.’

5. A typological comparison between Persian, English, Japanese, and Hebrew In this section, we will offer a cross-linguistic perspective of tense interpretation in past-under-past, present-underpast, and present-under-future constructions in the complement of attitude verbs and relative clauses of Persian, English, Japanese, and Hebrew. Because we will glean insight from Ogihara and Sharvit (2012) and attempt to incorporate Persian data into their inventory, we do not provide any examples from Japanese, and Hebrew and refer the interested readers to Ogihara and Sharvit (2012) for Japanese and Hebrew examples. As for past-under-past construal in complement clauses, as discussed above, such constructions in Persian can convey both the simultaneous past and the back-shifted readings. English past-under-past complement clauses express the simultaneous past reading and can encode the back-shifted reading, if an appropriate adverb is used. However, in Japanese, past-under-past complement clauses do not allow the simultaneous past reading and merely encode the backshifted reading. In Hebrew, past-under-past complement clauses are envisaged as encoding merely a back-shifted reading by most Hebrew speakers; while, few Hebrew speakers assume a simultaneous past reading for thusconstructions (Hatav, 2012; Ogihara and Sharvit, 2012). The interpretation of past-under-past constructions in relative clauses is similar in Persian, English, Japanese, and Hebrew in that all of them can convey the back-shifted, simultaneous, and forward-shifted readings.

12

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

Table 1 Possibility of encoding the simultaneous reading.

a

English complements of attitude verbs English relative clauses Hebrew complements of attitude verbs Hebrew relative clauses Japanese complement of attitude verbs Japanese relative clauses Persian complement of attitude verbs Persian relative clauses

Past-under-past

Present-under-past

Present-under-future

Possible Possible Often impossible Possible Impossible Possible Possible Possible

Impossible Impossible Possible Impossible Possible Possible Possible Impossible

Possible Possible Possible Possible Possible Possible Possible Possible

a ‘‘As discussed in Section 2, in this article, for tense interpretation in the complement clauses of Persian, both verbs of saying and verbs of attitude were examined. However, given that the table is borrowed from Ogihara and Sharvit (2012) and they were merely concerned with complements of verbs of attitude, tense interpretation in the complement of verbs of saying is not provided (However, as the interview results revealed, in Persian, the interpretation of tense in the complement of verbs of saying and verbs of attitude does not differ from each other in pastunder-past, preset-under-past, and present-under-future constructions). As for the relative clauses, we were not concerned with the aspectual class of the matrix clause tense and were merely concerned with the tense interpretation in relative clauses with different tenses. Hence, in the table, we did not specify any specific verb type with a specific aspect for tense interpretation in relative clauses.

With regard to present-under-past construal in complement clauses, it was revealed that, in Persian, thusconstructions convey the simultaneous past reading. However, in English, due to the indexical nature of its present tense, the embedded present tense in present-under-past complement clauses, expresses the DAR. In Japanese and Hebrew, as traditionally-regarded non-SOT languages, these constructions encode the simultaneous past reading. But, in Japanese, the embedded present tense can encode simultaneity with the actual UT (if appropriate adverb is used) (Ogihara and Sharvit, 2012). In present-under-past relative clauses, we observe that Persian displays a behavior which is akin to English and Hebrew and differs from Japanese because, in Persian, English, and Hebrew, the embedded present tense cannot convey simultaneity with the matrix clause past tense and merely conveys simultaneity with UT. However, in Japanese, the embedded present tense can convey simultaneity with the matrix clause past tense. Regarding present-under-future construal in complement clauses, we found that Persian, English, Japanese, and Hebrew allow simultaneity with the matrix future tense. In Persian, English, Japanese, and Hebrew, the embedded present tense of present-under-future relative clause, does not necessarily refer to the actual UT and expresses simultaneity with the matrix clause future tense. In Table 1, we examine the possibility of encoding simultaneity with the matrix tense in past-under-past, present-underpast, and present-under-future constructions of English, Hebrew, Japanese, and Persian. The in-depth analysis of the table highlights one of the other differences between English and Persian present tenses and provides further evidence for the non-indextical nature of Persian present tense. As can be observed, in English, the encoding of simultaneous reading by present tense depends on the matrix clause tense in that, when the matrix tense is past (e.g. example (30)), the embedded present tense must overlap the UT and this is revealed by the unacceptability of sentence (31). However, when the matrix tense is future (e.g. example (32)), the embedded present tense is not required to be simultaneous with the actual UT and it can be co-temporal with the matrix clause tense without overlapping with the UT. This is confirmed by the acceptability of the sentence (33). This claim is in line with the claim of Smith (1978), Hatav (2010), Sharvit (2003), etc. that the accuracy of present-under-past constructions in English and other languages, which have indexical present tense, is dependent on having the DAR. However, different rules govern the interpretation of Persian present tense in complement clauses in that, in Persian, the simultaneous reading of Persian present tense in complement clauses is not governed by the matrix clause tense. This argument is buttressed by examples (24) (presentunder-past in complement clause) and (26) (present-under-future in complement clause) in that, in neither of them, the time of Maryam's loving must overlap the UT. This is demonstrated by the acceptability of the sentences in (14) and (27), respectively. However, in relative clauses with present tense, having a simultaneous reading or not is determined by the matrix clause tense such that the present tense encodes the simultaneous reading if the matrix clause is in the future but is devoid of the simultaneous reading if it is in the past. In fact, in example (25) with a present tense in a relative clause embedded under a past tense, the crying time must be simultaneous with the UT. While, in (28) with a present tense in a relative clause embedded under a future tense, the loving time can be simultaneous with the UT, but need not and it is demonstrated by (29). 30) John found out that Mary is pregnant.

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

13

31) *Mary is not pregnant now but she was once and John found out that Mary is pregnant. 32) John will find out that Mary is pregnant. 33) Mary is not pregnant now but she will become in the following three months and John will find out that Mary is pregnant.

6. Persian tense system: SOT or non-SOT In this section, we will investigate whether Persian is an SOT or non-SOT language. To achieve this aim, in Section 6.1, we will provide two pieces of evidence for regarding Persian as an SOT language. Then, in Section 6.2, we will challenge the augments for regarding Persian as an SOT language and provide five arguments supporting the non-SOT nature of Persian. Finally, in Section 6.3, we will provide some of evidence which undermine the arguments provided in Sections 6.1 and 6.2 and prove the claim that Persian is a hybrid language. 6.1. Reasons for regarding Persian as an SOT language As mentioned above, one of the primary criteria for distinguishing SOT languages from non-SOT languages is whether, in past-under-past constructions, the embedded tense can be construed as a null tense giving rise to a simultaneous past reading (Ogihara and Sharvit, 2012; Sharvit, 2014). Using this criterion as our point of reference, we come to the conclusion that Persian, likewise English, should be subsumed under the rubric of SOT languages. Because, as interview results revealed, Persian past-under-past constructions, irrespective of the verb of the matrix clause (both with saying and attitudinal matrix verbs), have the option of receiving a simultaneous past reading, in addition to the back-shifted reading (refer to examples 3--6). Another observation verifying the claim that Persian falls among SOT languages is that some of its characteristics are like Modern Greek which is considered as an SOT language by Sharvit (2003). In Modern Greek, the simultaneous past reading is expressed using both present and past tenses (Sharvit, 2003, 2014; Schlenker, 1999). Accordingly, it is considered as a language with both an SOT rule and a shiftable present tense (Sharvit, 2003, 2014). As shown by the second part of the interview, in Persian complement clauses, likewise Modern Greek, simultaneity with matrix clause past tense can be expressed either using a present or a past tense (refer to examples 7--10). The observed similarity between Persian and Modern Greek provides us with cogent argument to claim that Persian can also be construed as an SOT language with a shiftable present tense. Over-reliance on theories regarding SOT as a rule accounting for the simultaneous past reading of past-under-past constructions leads to categorizing Persian as an SOT language. However, in Section 6.2, in light of the counterexamples against considering Persian as an SOT language, we propose a modification and provide some pieces of evidence for regarding Persian as a non-SOT language.

6.2. Reasons for regarding Persian as a non-SOT language Firstly, one of the arguments challenging the claim that Persian falls among SOT languages is based on Sharvit's (2014) position that, although some languages, which are traditionally construed as non-SOT languages, allow a simultaneous past reading in the counterpart of examples such as 1 (repeated in 34), none of these languages allow the simultaneous past reading in the counterpart of example (35) in which neither ‘‘singing’’ nor ‘‘arrival’’ precedes the UT but a past tense is used: 34) John said that Mary was pregnant. 35) John said that Mary would sing before she arrived. As illustrated by the interview questions, in Persian, the equivalent of example (34) (i.e., example 3) allowes simultaneous past reading; however, in Persian, in the equivalent of example (35) provided below (example (36)), a future tense; instead of a past tense, is used. Hence, the possibility of encoding a simultaneous past reading is eliminated.

14

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

36) ʔæli goft qæble æz inke ʔu be-res-e, mæryæm xah-æd xand Ali say-PAST before from that he SUB-arrive-3SG Maryam will-3SG sing-PAST ‘Ali said that Maryam would sing before he arrived.’ Secondly, the other counterexample against considering Persian as an SOT language is that, in Persian, past-underpast constructions with an intervening future-oriented adverb are devoid of the simultaneous past reading and merely encode the back-shifted reading. This is instantiated by example (37) in which a past tense is embedded under another past tense with a future-oriented adverb intervening between them. In this sentence, the time of ‘‘missing’’ in the most embedded clause cannot be construed as being simultaneous with the time of ‘‘saying’’ in the immediately c-commanding clause; rather, it is construed as preceding the time of ‘‘saying’’. Our claim is substantiated by one of the interview questions, in which, to express simultaneity with matrix clause past tense with an intervening future-oriented infinitive, 31 of the participants concurred on using the present tense (example 38). 37) hæfteye piš, ʔæli fekr kærd ke 10 ruze ʔayænde be maman-eš xah-æd week ago, Ali think do-PAST that 10 day next to mother-his will-3SG goft ke del-eš bær-aš tæng šod-e bud. say-PAST that heart-his for-her small become-PSP be-PAST. ‘Ali thought a week ago that in 10 days he would tell his mother that he had missed her.’ 38) hæfteye piš, ʔæli fekr kærd week ago, Ali think do-PAST goft ke del-æš bær-aš tæng say-PAST that heart-his for-her small ‘Ali thought a week ago that in 10 days he

ke 10 ruze ʔayænde be maman-eš xah-æd that 10 day next to mother-his will-3SG šode ʔæst. become be-PRESENT. would tell his mother that he missed her.’

Thirdly, by adopting Grønn and von Stechew's (2010) conception of SOT as our criterion, we come to the conclusion that Persian can be categorized as a non-SOT language. The reason is that, according to Grønn and von Stechew (2010: 109), in SOT languages, there is an ‘‘agreement’’ between the matrix and the embedded clause tenses. In fact, the authors refer to SOT and non-SOT languages as ‘‘tense agreement’’ and ‘‘tense nonagreement’’ languages, respectively. Zagona (2013), in the same line of thought, asserts that SOT clauses have an ‘‘interpretable present’’ tense and the past form is the manifestation of an ‘‘agreement’’ with the matrix clause past tense. In Persian, there is no rule requiring the embedded tense to agree with the matrix clause tense. This was made manifest in transforming direct speech into indirect speech in the third part of the interview (example 11). This finding is also confirmed by Mace (1971) and Bakhtiyari's (2001) argument that, in Persian, no distinction can be drawn between direct and indirect speech in terms of their tenses. For example, in examples (39) and (40), although the matrix clause is in the past, the tenses of the verbs used in the embedded clauses of indirect speech sentences are identical with the tenses in the direct speech sentences and they are unaffected by the matrix clause tense. ʔæli goft: ‘‘mæryæm bardar ʔæst’’ Ali say-PAST: ‘‘Maryam pregnant be-PRESENT’’ ‘Ali said ‘‘Maryam is pregnant’’’ Indirect speech: ʔæli goft ke mæryæm bardar ʔæst. Ali say-PAST that Maryam pregnant be-PRESENT. ‘Ali said that Maryam is/was pregnant.’

39) Direct speech:

dust dar-æm’’ 40) Direct speech: ʔæli goft: ‘‘mæryæm- ra Ali say-PAST: ‘‘Maryam-ACC love have-1SG’’ ‘Ali said ‘‘I love Maryam’’’ Indirect speech: ʔæli goft (ke) ʔu mæryæm-ra dust dar-æd. Ali say-PAST (that) he Maryam-ACC love have-3SG ‘Ali said that he loves/loved Maryam.’ Thus, in Persian, we cannot argue that (41a) is the indirect speech of (41b) and an SOT rule is applied to transform the embedded present tense of direct speech into a past tense in the indirect speech.

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

15

41a) ʔæli goft mæryæm mæriz bud Ali say-PAST Maryam sick was-PAST ‘Ali said that Maryam was sick.’ ‘‘mæryæm mæriz ʔæst’’. 41b) ʔæli goft Ali say-PAST ‘‘Maryam sick be-PRESENT’’. ‘Ali said that Maryam is sick.’ Fourthly, the argument that Persian is a non-SOT language is further confirmed when considering the embeddibility principle, as proposed by Sharvit (2003, 2014). Based on the embeddibility principle, every language can express the following generic belief using embedding: Three hundred years ago, Mary had the following belief: ‘‘two is not an even number’’. According to Sharvit (2014), in SOT languages, such generic beliefs are expressed using a past-under-past construction. Because the SOT rule is applied to the embedded past tense and the intended interpretation is obtained. In fact, the present tense of these languages is indexical; thus, it cannot merely express simultaneity with the matrix clause past tense (without referring to the UT). However, in languages with shiftable present (a present tense not required to refer to the actual UT in present-under-past constructions), it is expressed using a present-under-past construction. Given that, in Persian, Maryam's generic belief is expressed using a present tense (example 42), we can argue that Persian cannot be construed as an SOT language. This line of reasoning also provides convincing evidence for the non-indexical nature of Persian present tense. 42) 300 sal piš, mæryæm bavær dašt ke 2 zoj ni-st. 300 year ago, Maryam believe have-PAST that 2 even NEG-be-PRESENT ‘300 years ago, Maryam believed that 2 was not an even number.’ Finally, the other evidence for the non-SOT nature of Persian comes from the ungrammaticality of using past tense in Persian before-clauses (refer to example (19)) because Sharvit (2014) maintains that, in languages with quantificational past tense,8 the use of past tense in before- clauses leads to an ill-formed sentence or, to put it more precisely, ‘‘no language (that lacks an SOT rule) can have a quantificational past tense in a before-clause’’ (Sharvit, 2014: 265). Since the use of past tenses in Persian before-clauses leads to ungrammaticality, we can conclude that not only Persian past tense is quantificational (rather than pronominal) but also it lacks the SOT rule. In fact, if Persian past tense was pronominal or it had the SOT rule, the use of past tense in its before-clauses did not lead to ill-formed sentences. In this section, five arguments were provided to substantiate the claim that Persian is a non-SOT language and challenge the arguments provided in Section 6.1 for considering Persian as an SOT language. In Section 6.3, we will provide some arguments for considering Persian as a hybrid language exhibiting some of the characteristics of SOT and some of the characteristics of non-SOT languages.

6.3. Reasons for regarding Persian as a hybrid language Firstly, by adopting the criteria Sharvit (2003: 669) laid down for distinguishing SOT languages from non-SOT languages, we come to the conclusion that Persian is a hybrid language defying the clear-cut SOT/non-SOT classification. These criteria are as follows: in SOT languages, (i) the embedded past of past-under-past constructions is construed as being semantically non-past and (ii) present-under-past constructions may only encode DAR. While, in nonSOT languages, (i) present-under-past constructions have a non-past reading corresponding to the non-past reading of past-under-past constructions in SOT languages and (ii) their past-under-past constructions merely encode the anteriority or back-shifted reading. Applying Sharvit's (2003) criteria on Persian data, we come to the conclusion that Persian, on account of allowing the simultaneous past reading of past-under-past constructions, fulfills the first criteria of SOT languages and proves to have some of the properties of SOT languages. However, because Persian present-under-past constructions do not have merely the DAR, it does not satisfy the second criteria of SOT languages. As for the criteria of non-SOT languages, because Persian present-under-past constructions (in tandem with past-under-past constructions) can be used for the expression of simultaneous past, it fulfills the first criteria of non-SOT languages. However, since, in Persian, past-under-

8 A quantificational past tense is an existential quantifier over times and using it in before-clauses is ill-formed because when p (in q is before p) includes a quantificational past tense, the embedded past tense is an existential quantifier over times and a unique reference point cannot be selected (this is referred to as presupposition failure). In fact, in the sentence ‘‘I sang before John played’’, we cannot find a unique time of John's playing in that, for every t in which John plays, there is a t’ before t such that John's playing is before t’ (Sharvit, 2014).

16

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

past constructions do not merely encode the back-shifted reading and express the simultaneous past, Persian fails to satisfy the second criteria of non-SOT languages. The inconsistency observed in Persian behavior may justify arguing that Persian is a hybrid language. Secondly, the claim that Persian is a hybrid language is further confirmed by considering approaches dealing with the ties between the matrix indexicality of present tense and the SOT phenomenon. For example, the main argument of Ogihara (1999) and Hatav's (2012) approach for distinguishing SOT languages from non-SOT languages is that, for the expression of simultaneous past, if the present tense can be used under a matrix clause past tense without being required to overlap the UT, that language is a non-SOT language. However, if the present tense is indexical and must refer to the actual UT, it cannot be used for the purpose of expressing simultaneity with the matrix clause past tense. Hence, a past tense is used instead and the SOT rule is used to delete the embedded past tense and obtain the intended meaning. The interrelationship between the indexicality of present tense and the SOT phenomenon also constitutes the cornerstone of Sharvit's (2003, 2014) embeddibility principle. The central tenet underlying Sharvit's argument is that if present tense in a language is obligatorily indexical and necessarily refers to the actual UT, that language allows a null or vacuous past in past-under-past constructions. Delving deeper into Persian data, we find that Persian data belie this argument. The reason is that, as demonstrated in previous sections, Persian present tense is not indexical and is not necessarily required to refer to the actual UT in present-under-past complement clauses but it has a null or vacuous past. Because, as confirmed by interview results, past-under-past constructions allow the non-past or simultaneous past reading. In fact, in Persian, a present tense can be used for the expression of simultaneous past. Thus, the view that Persian present tense is not indexical and; accordingly, Persian is a non-SOT language gains ground. However, in Persian, past-under-past constructions can also be used for the expression of simultaneous past. Hence, we might claim that an SOT rule is required to delete the embedded past tense. The observed inconsistency in Persian data is resolved by arguing that, Persian is a hybrid language defying the SOT/non-SOT classification. Thirdly, considerable insights into the hybrid nature of Persian can be gained by taking ‘‘relative’’ clauses into account. In fact, in Persian ‘‘relative’’ clauses, past-under-past constructions express simultaneous past (example 23) and presentunder-past clauses (example 25) merely encode simultaneity with UT (refer to Sections 4.1 and 4.2). While, in ‘‘complement’’ clauses, both present-under-past and past-under-past constructions can be used for the expression of simultaneous past. Accordingly, Persian relative and complement clauses do not exhibit a consistent behavior because, in relative clauses, Persian displays SOT-like behaviors; however, in complement clauses, it has some of the properties of SOT and some of the properties of non-SOT languages. However, the observations regarding tense construal in Persian complement and relative clauses are not difficult to reconcile in that they reveal that the distinction between SOT and non-SOT languages is blurred and a clear-cut distinction cannot be made between them. This conclusion confirms Kusumoto (1998) claim that the classification of languages into SOT and non-SOT languages is not on the right track because Polish and Russian exhibit the behavior of SOT languages in their complement clauses; while, in their relative clauses, they behave like non-SOT languages. Furthermore, Ogihara and Sharvit (2012) categorize English as a SOT language and Japanese as a non-SOT language. However, being informed that Hebrew has the properties of both SOT (in relative clause) and non-SOT languages (in complement clauses), they assert that drawing a clear-cut distinction between SOT and non-SOT languages is untenable. The claim that Persian is a hybrid language reveals important information regarding the overall grammatical structure of Persian. Firstly, it explicates why, in Persian complement clauses, a present tense is allowed under a matrix clause past tense without being obliged to encode simultaneity with the UT. Secondly, it accounts for why, in Persian, both presentunder-past and past-under-past constructions are amenable to the simultaneous past reading. Finally, it justifies why, in past-under-past complement clauses, although no tense shift is applied to transform the direct speech into an indirect speech, they convey a simultaneous past reading. Having argued that Persian is a hybrid language, in the next section, we will investigate how the simultaneous past reading of its past-under-past constructions can be accounted for. This is a very important issue because the analyses which draw on the SOT rule to obtain the simultaneous past reading cannot be adopted for Persian which is a hybrid language.

7. The analysis of past-under-past in Persian complement clauses In the SOT literature, one of the important issues is accounting for the simultaneous past reading of some of pastunder-past constructions in non-SOT languages. This issue is important because these languages, as their name indicates, are characterized by lacking an SOT rule to account for or obtain the simultaneous past reading. Ogihara and Sharvit (2012) propose a ‘‘de re’’ interpretation to capture the simultaneous past reading of past-under-past constructions in Hebrew, a traditionally-construed non-SOT language. Grønn and von Stechow (2010), likewise, attempted to account for the simultaneous past reading of Russian past-under-past constructions with a factive verb in the matrix clause by the

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

17

de re approach. The other proposal was made by Hatav (2012), who referred to interval semantics to account for the simultaneous past reading of some of past-under-past constructions in Hebrew. In a rather similar line of thought, Gennari (2003: 56) states that the simultaneous past reading of past-under-past constructions in SOT languages such as English can be captured by arguing that the interval denoted by the embedded clause past tense can extend from an interval earlier than the matrix clause past interval until the matrix clause interval itself. Given that Persian is a hybrid language, the analyses proposed for SOT languages, which draw on the SOT rule (irrespective of its definition) to account for the simultaneous past reading of past-under-past constructions, cannot be adopted. Thus, the analyses which do not appeal to the SOT rule (such as the four analyses mentioned above) fit the Persian data better. Gennari's pragmatic-driven (2003) and Hatav's semantic-driven (2012) approaches are not adopted because they fail to capture the cross-linguistic variation of past-under-past constructions and explicate why, in some languages, past-under-past constructions do not lend themselves to the simultaneous past reading; while, in some languages, they do. Grønn and von Stechow's (2010) analysis, likewise, is not deployed in that, it is only proposed for factive verbs; while, in Persian, attitudinal, saying, and factive verbs allow the simultaneous past reading. Thus, to account for the simultaneous past reading of past-under-past complement clauses in Persian, as a hybrid language, we use the ‘‘de re’’ theory, as proposed by Abusch (1997) and further developed by Ogihara and Sharvit (2012). The de re theory is rooted in the philosophy of language and is proposed to account for the interpretation of nominal expressions in belief and intentional contexts. However, since, tenses are treated on a par with pronouns (Partee, 1973), the de re theory can be borrowed and adopted for tense construal in belief and intentional contexts. One of the basic tenets of the de re theory is the ‘‘res movement’’ which is defined as follows: when referential expressions are used in intentional contexts such as the complement of attitude verbs, they may (perhaps must) move out of the complement clause to an argument position of the attitude verb in the matrix clause (Heim, 1994:154). Hence, in embedded tense construal, a tense (as a referential expression) which is embedded in an attitude complement (and is referred to as the res) may or must move out of the complement to an argument position of the matrix clause attitude verb. Besides the res movement, the other assumption of the de re analysis is the acquaintance relation which determines the denotation of res (here, the embedded tense) outside the attitude context (Ogihara and Sharvit, 2012: 650). In fact, the acquaintance relation, as its name suggests, refers to the way the attitude holder was acquainted with the state or event in the embedded clause (i.e., the res) or it relates an individual to the res. One of the other major premises of the de re theory is that the reference or the trace of an embedded tense should not be after the attitude holder's now (Abusch, 1997). This is referred to as the upper limit condition (ULC). This is a very simplified sketch of the de re theory provided for the purpose of the present study. Drawing on the above premises, Ogihara and Sharvit (2012) account for the simultaneous past and back-shifted readings of past-under-past constructions in non-SOT languages. In the following, we adopt this line of reasoning to account for the simultaneous and back-shifted readings of Persian past-under-past complement clauses. 43) Mæn bavær dašt-æm ke mæryæm mæn-ro dust dašt. I believe have-PAST-1SG that Maryam I-ACC love have-PAST I believed that Maryam loved me. [Mæn PAST0,2< [bavær dašt-æm DE-REPAST<0,3] l3l0[mæryæm [e3- mæn-ro dust dašt]]]. Before explicating the details, two notational remarks regarding the above representation should be made. In this theory, because undeleted tenses require two times for their interpretation (namely evaluation time and event time), they are doubly-indexed, with the first index denoting the local evaluation time and the second index denoting the event time. In fact, undeleted tenses are like complex pronominal expressions. Furthermore, the < feature, representing past tense, encodes anteriority. In the de re analysis of the above past-under-past construction, at first, due to the res movement, the embedded past tense moves out of the complement clause. Then, due to the ULC requirement, the trace of the embedded past tense cannot be posterior to the attitude holder's now. Hence, the adopted acquaintance relation, relating the res (i.e., Maryam's loving me) to the attitude holder (i.e., I), is either simultaneous with the attitude holder's now or is anterior to it. If the selected acquaintance relation is identity with attitude holder's now, the ULC is satisfied and the simultaneous reading of this sentence is obtained. However, if the adopted acquaintance relation is before the attitude holder's now (for example, ‘‘in my childhood’’), again the ULC is fulfilled and the back-shifted reading is accounted for. As can be observed, due to the hybrid nature of Persian, the ‘‘de re’’ analysis which does not appeal to the SOT rule is used for the analysis of Persian data. 8. Conclusion This article investigated tense construal in Persian past-under-past, present-under-past, and present-underfuture complement and relative clauses. However, because the researchers did not converge on the tense construal of

18

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

past-under-past constructions in complement clauses (which is one of the criteria for distinguishing SOT languages from non-SOT languages), 32 Persian native speakers were interviewed to reach a consensus. The results of the interview demonstrate that, in Persian, irrespective of whether the verb of the matrix clause is attitudinal or saying, both presentunder-past and past-under-past constructions can be used for the expression of simultaneous past. Furthermore, this study examined whether Persian is an SOT or non-SOT language. In answering this question, we were saddled with determining the nature of Persian present tense because the nature of present tense and the SOT phenomenon are tied. Persian present tense was revealed to be non-indexical because it is not required to refer to the actual UT in presentunder-past complement clauses. Then, it was argued that Persian is a hybrid language challenging the well-established SOT/non-SOT distinction. The reason is that, in complement clauses, it displayed a behavior which was both like SOT languages (such as using past-under-past constructions for the expression of simultaneous past) and non-SOT languages (such as using present-under-past for the expression of simultaneous past). In relative clauses, however, its behavior was akin to SOT languages because only past-under-past constructions are used for the expression of simultaneous past. Finally, the de re analysis, as proposed by Abusch (1997) and further developed by Ogihara and Sharvit (2012), was used to account for the simultaneous past reading of past-under-past constructions in Persian, as a hybrid language. In fact, because Persian was a hybrid language, SOT-based analyses could not be adopted to capture the simultaneous past reading of its past-under-past constructions. References Abusch, D., 1997. Sequence of tense and temporal de re. Linguist. Philos. 20, 1--50. Altshuler, D., 2004. A simultaneous perception of things: SOT in Russian. Snippet 8, 5--6. Altshuler, D., 2008. Narrative progression effects in Russian indirect reports and what they reveal about the meaning of the past tense’. In: Friedman, T., Ito, S. (Eds.), Proceedings of Semantics and Linguistic Theory, vol. 18. CLC Publications, Cornell University, pp. 19--36. Altshuler, D., Schwarzschild, R., 2013a. Correlating cessation with double access. In: Aloni, M., Franke, M., Roelofsen, F. (Eds.), Proceedings of the 19th Amsterdam Colloquium. ILLC, University of Amsterdam, Amsterdam, pp. 43--50. Altshuler, D., Schwarzschild, R., 2013b. Moments of change, cessation implicatures and simultaneous readings. In: Chemla, E., Homer, V., Winterstein, G. (Eds.), Proceedings of Sinn und Bedeutung, vol. 17. ENS, Paris, pp. 45--62. Bakhtiyari, B., 2001. Sequence of tenses in Persian. In: Guertsenberg, L. (Ed.), Proceedings of the Celebration of the 110th Anniversary of Professor Viktor M. Zhimunsky, pp. 122--138. Bary, C., Altshuler, D., 2014. Double access. In: Csipak, E., Zeijlstra, H. (Eds.), Proceedings of Sinn und Bedeutung, vol. 19. ENS, Paris, pp. 89-106. Boogaart, R., 1999. Aspect and temporal ordering: a contrastive analysis of Dutch and English (Ph.D. dissertation). Free University of Amsterdam. Enç, M., 1987. Anchoring conditions for tense. Linguist. Inquiry 18, 633--657. Gennari, S., 2003. Tense meanings and temporal interpretation. J. Seman. 20, 35--71. Grønn, A., von Stechow, A., 2010. Complement tense in contrast: the SOT parameter in Russian and English. In: Grønn, A., Marijanovic, I. (Eds.), Russian in Contrast. Grammar. Universiteteti Oslo, Oslo, pp. 109--154. Grønn, A., von Stechow, A., 2011. Tense in subjunctive conditionals across languages. Ms. University of Oslo. Grønn, A., von Stechow, A., 2013. Tense in adjuncts. Lang. Linguist. Compass 7 (5), 311--327. Haegeman, L., 2008. Introduction to Government and Binding Theory, 2nd ed. Basil Blackwell, Oxford. Hatav, G., 2010. Relative and absolute tense interpretation in Modern Hebrew. Hebrew Stud. 51, 261--285. Hatav, G., 2012. Bound tenses. In: Binnick, R. (Ed.), The Oxford Handbook of Tense and Aspect. Oxford University Press, Oxford, pp. 611--637. Heim, I., 1994. Comments on Abusch's theory of tense. In: Kamp, H. (Ed.), Ellipsis, Tense and Questions, Dyana-2 Esprit Basic Research Project 6852, Deliverable R2.2.B. Manuscript. MIT. Kaplan, D., 1977. Demonstratives. Ms. UCLA, Los Angeles, CA. Khomitsevich, O., 2007. Dependencies across phases: from sequence of tense to restrictions on movement (Ph.D. dissertation). Utrecht University. Kusumoto, K., 1998. Typology of Embedded Tenses in Non-SOT Languages. A Talk given at Workshop on Syntax and Semantics of Tense. Yokohama National University. Mace, J., 1971. Modern Persian. English Universities Press, London. Ogihara, T., 1995a. The semantics of tense in embedded clauses. Linguist. Inquiry 26, 663--679. Ogihara, T., 1995b. Double-access sentences and reference to states. Nat. Lang. Seman. 3, 177--210. Ogihara, T., 1996. Tense, Attitudes, and Scope. Kluwer Academic Publishers, Dordrecht and New York. Ogihara, T., 1999. Double-access sentences generalized. In: Proceedings from Semantics and Linguistic Theory IX (SALT9). CLC Publications, Cornell University, pp. 224--236. Ogihara, T., 1994. Adverbs of quantification and sequence-of-tense phenomena. In: Santelmann, L., Harveys, M. (Eds.), Proceedings of Semantics and Linguistic Theory. CLC Publications, Ithaca, NY, Cornell University, pp. 251--267. Ogihara, T., Sharvit, Y., 2012. Embedded tenses. In: Binnick, R. (Ed.), The Oxford Handbook of Tense and Aspect. Oxford University Press, Oxford, pp. 638--668. Partee, B., 1973. Some structural analogies between tenses and pronouns in English. J. Philos. 70, 601--609. Radford, A., 2006. Minimalist Syntax: Exploring the Structure of English. Cambridge University Press, Cambridge. Schlenker, P., 1999. Propositional attitudes and indexicality: a cross-categorical approach (Ph.D. dissertation). MIT, Cambridge, MA. Schlenker, P., 2003. A plea for monsters. Linguist. Philos. 26, 29--120. Sharvit, Y., 2003. Embedded tense and universal grammar. Linguist. Inquiry 34, 669--681.

M. Sameri, G. Karimi-Doostan / Lingua 226 (2019) 1--19

19

Sharvit, Y., 2014. On the universal principles of tense embedding: the lesson from before. J. Seman. 31 (2), 263--313. Sharvit, Y., 2018. Sequence of tense: syntax, semantics, and pragmatics. In: Potel-Grosz, P., Georg Grosz, P., Zobel, S. (Eds.), Pronouns in Embedded Contexts at the Syntax-Semantics Interface. Springer, USA, pp. 215--247. Smith, C., 1978. The syntax and interpretation of temporal expressions in English. Linguist. Philos. 2 (1), 43--100. Stowell, T., 2007. The syntactic expression of tense. Lingua 117, 437--463. Zagona, K., 2004a. Tense construal in complement clauses: verbs of communication and the double access reading. In: Gueron, J., Lecarme, J. (Eds.), The Syntax of Time. The MIT Press, Cambridge, MA, London, England, pp. 637--653. Zagona, K., 2004b. Measuring out complement clause tense. Probus 16, 273--315. Zagona, K., 2005. The spatio-temporal path and aspectual composition. In: Kempchinsky, P., Slabakova, R. (Eds.), Aspectual Inquiries. Springer, Dordrecht, pp. 171--190. Zagona, K., 2013. Tense, aspect and modality. In: den Dikken, M. (Ed.), Encyclopedia of Generative Syntax. Cambridge University Press, Cambridge, pp. 746--792.