Deixis and Embedded Tense: Revisiting tense in English and Japanese subordinate clauses

Emilia Melara University of Toronto

This investigation examines tense in English and Japanese, specifically how tense in embedded clauses interacts with matrix clause tenses. Sequence of tense (henceforth SOT) traditionally refers to the pattern observed of some languages in which temporal coincidence between an embedded clause and a past matrix clause is expressed by a subordinate past tense (Enç 1987; Ogihara 1996; Stowell 2007). Analyses of tense vary between referential and quantificational approaches. I propose a treatment in which the tense morphology that surfaces in embedded clauses is a function of the ways in which they are anchored with respect to time and point of view. English embedded tenses are interpreted relative to the utterance time. Japanese embedded tenses, conversely, are interpreted relative to the time of their embedding clauses. English uses the past-underpast configuration to express coincidence between matrix and embedded situations (I use “situation” to refer to events and states together), for example, Becky said that Kathryn was hungry (i.e. at the time of saying). The past-under-past sequence also derives a past-shifted reading, under which the embedded situation precedes the matrix situation (e.g. Becky said a moment ago that Kathryn was hungry earlier). In past relative clauses under a matrix past, a forward-shifted reading may also arise, under which the embedded situation holds between the matrix time and utterance time (e.g. (Last week,) Becky was looking for a man who was on TV (this morning)). Past-under-past constructions in Japanese produce only the past-shifted reading. Japanese uses the present-under-past configuration to express simultaneity between embedded and embedding clauses. Thus, the complement clause in (1) and the relative clause in (2) are interpreted as coinciding with the time of their corresponding matrix situations. (1) Hanako-wa Akira-ga hon-o yon-de i-ru to it-ta Hanako-TOP Akira-NOM book-ACC read-TE be-PRES that say-PST ‘Hanako said that Akira was reading a book.’ (i.e. Hanako said: “Akira is reading a book.”) (2) Taroo-wa [nai-te i-ru otoko]-o mi-ta Taro-TOP cry-PROG be-PRES man-ACC see-PST ‘Taro saw a man who was crying (at the time of the meeting).’ (Ogihara 1996:154) Present relative clauses under past matrix verbs in English (e.g. Becky was looking for a man who is on TV) are, rather, necessarily interpreted as holding at the utterance time. Unlike Japanese also, an English present complement under a matrix past (e.g. Laura said that Becky is pregnant) is interpreted as holding at both the time of the matrix situation and the utterance time – the double-access reading. Like English, however, in Japanese a past relative clause under a matrix past, as in (3), can receive the forward-shifted interpretation (Enç 1987). In both languages, the forward-shifted reading is possible only when the NP modified by the relative clause is read de re (“of the thing”) and therefore not when read de dicto (“of the saying”) (Abusch 1988). (3) Taroo-wa [[nooberu-syoo-o tot-ta] otoko]-o sagasi-ta Taro-TOP Nobel-prize-ACC win-PST man-ACC seek-PST ‘Taro looked for a/the man who won a Nobel prize.’ (Ogihara 1996:159) I propose what distinguishes SOT and non-SOT languages is how clauses are anchored temporally. Following Cowper’s (2005) feature geometry for the Infl domain, I assume the head of TP may have tense and temporal anchoring features. The past tense arises from the tense feature [Precedence], which sets at least one moment of a situation prior to a temporal anchor. Present tense morphology is unspecified for tense features. The feature [T-deixis] anchors a clause to the utterance time. In my analysis, clauses in the indicative mood must be anchored temporally, ultimately to the moment of speech. A temporal anchor is a time relative to which another time is evaluated by means of a tense feature. The time evaluated is a reference point: the time at which the situation denoted by vP holds. Not all languages have T-deixis. If a language has T-deixis in its feature inventory, it is specified on indicative clauses. What distinguishes the English and Japanese tense systems, with respect to SOT, is that English has T-deixis, while Japanese lacks it. Because indicative Ts must be anchored temporally, indicative Ts in English are T-deictic – a situation’s time is interpreted relative to the utterance time. In Japanese, on the other hand, matrix Ts are anchored to the moment of speech by default and each lower clause is anchored by the reference point of the embedding clause. Indicative clauses must also be anchored to a point of view, which is achieved with Cowper’s [Pdeixis]. This feature anchors a clause to the speaker’s perspective and, for me, is represented in the head of CP. In both English and Japanese, attitude verbs s-select non-P-deictic clauses. Because 1

clauses must be anchored to a point of view, the lower clause gets anchored to the point of view of the matrix subject, bearing the agent/experiencer theta-role. When P-deixis is specified, as it may be in relative clauses, the forward-shifted reading arises and an NP may be read de re. Thus, we capture why the forward-shifted reading is present in constructions with relative but not complement clauses. Previous accounts invariably treat the past-shifted and simultaneous readings of past-under-past constructions as distinct semantically and syntactically. I treat the two as a single reading: before utterance time. In English past-under-past constructions, T-deixis on each T ensures that both matrix and subordinate situations precede the utterance time. The lack of P-deixis on an embedded clause ensures that the lower situation is not read as holding after the matrix subject’s past. In relative clauses, the optional presence of P-deixis may allow a forward-shifted interpretation. In Japanese pastunder-past constructions, as T cannot be T-deictic, the lower clause is anchored by the matrix reference point, which precedes the utterance time. Precedence on the lower T sets the clause’s reference point to a time prior to that of the matrix clause and the past-shifted reading arises. Deriving simultaneity in Japanese therefore requires the present-under-past sequence as the absence of any tense features on the lower T means simultaneity between the anchor – the matrix clause’s reference point – and the lower clause’s reference point. Ogihara (1996) proposes that the difference between Japanese and English lies in whether the language has an SOT rule or not. This SOT rule, for him present in English but absent in Japanese, optionally deletes a lower tense at LF if c-commanded by an occurrence of the same tense. His (1999) account postulates that present tenses interpreted relative to the matrix time are actually underlyingly tenseless. Stowell (2007) proposes an account that treats past and present tense morphology as polarity and anti-polarity items, respectively. For him, semantic tense features appear on T, but tense morphology appears on a head lower than T. Past morphology, as PAST polarity items, may appear when either PAST or PRESENT semantic features are specified on a higher T, as long as there is a ccommanding PAST somewhere in the structure. Present tense morphology, conversely, as PAST antipolarity items, cannot appear under a c-commanding PAST semantic feature. Von Stechow & Grønn (2013a,b; also Grønn & von Stechow 2010) provide an analysis according to which the difference between SOT and non-SOT languages is that tense features are transmitted through a binding chain between a higher and lower clause in SOT languages, but this feature transmission does not hold in non-SOT languages. The T head is, for von Stechow & Grønn, comprised of a deictic, a bound, or a semantically vacuous pronoun and one of three tense relations. It is the relation between the pronominal forms in matrix and embedded clauses along with the presence or absence of the feature transmission parameter that determine how embedded clauses are interpreted. Under Ogihara’s account, the double-access reading arises in English because the English present tense morpheme is “absolute” and therefore inherently interpreted relative to the utterance time. Being an absolute tense, for Ogihara, an embedded present tense must be interpreted at both the “now” of the speaker and the “now” of the matrix clause’s subject. However, he does not make clear how the subject’s “now” enters the computation. The double-access reading, for him, does not arise in Japanese as the language lacks an absolute present. In Stowell (2007), it is never clear how tense morphemes are polarity items and, moreover, how they are semantically vacuous yet polarity items of semantic tense features. Von Stechow & Grønn’s approach, too, suffers for similar reasons. Although they do not treat tense morphemes as polarity items, they too consider tense morphology as semantically vacuous, licensed by semantic tense features higher in the clause. I show that their account is also incapable of accounting for the double-access reading. The present proposal provides a simpler solution to analysing tense cross-linguistically than previous analyses have allowed, offering a new perspective on inter- and intra-clausal inflectional relations. References Abusch, D. 1988. ‘Sequence of tense, intensionality, and scope.’ In Proceedings of WCCFL 7, 1–14. Cowper, E. 2005. ‘The Geometry of Interpretable Features: INFL in English and Spanish’. In Language 81(1):10-46. Enç, M. 1987. ‘Anchoring Conditions for Tense.’ In LI 18(4):633–657. Grønn, A. & A. von Stechow. 2010. ‘Complement Tense in Contrast: The SOT Parameter in Russian and English.’ In Russian in Contrast, Oslo Studies in Language 2(1). 109–153. Ogihara, T. 1996. Tense, Attitudes, and Scope. Kluwer. Ogihara,T. 1999. ‘Double-Access Sentences Generalized.’ In Proceedings of SALT9, 224-236. Stowell, T. 2007. ‘The syntactic expression of tense.’ In Lingua 117:437-463. von Stechow, A. & A. Grønn. 2013a. ‘Tense in Adjuncts Part 1: Relative Clauses.’ In Language and Linguistics Compass 7(5):295–310. --- 2013b. ‘Tense in Adjuncts Part 2: Temporal Adverbial Clauses.’ In Language and Linguistics Compass 7(5):311–327.

2