1 Introduction

In Japanese, the use of a negative preterite (also called “past perfective” or “simple past”) clause is discourse-pragmatically constrained, and oftentimes a negative nonpast-nonperfective (present-nonperfective) clause with -te iruFootnote 1 is used where a preterite clause is expected (Matsuda 2002, Yamashita 2004, Kusumoto 2016).

To illustrate, the preterite in (1B\(_a\)) sounds unnatural, conveying something to the effect that the speaker could have hired a new nurse; a nonpast-nonperfective clause is not pragmatically loaded in the same way, as seen in (1B\(_b\)) (the initial vowel of the auxiliary iru is often dropped in colloquial speech).Footnote 2

figure a

(2) illustrates that a negative preterite can be naturally used in a context where it was previously plausible (from the interlocutors’ viewpoint) that the logical contradiction of the propositional content would hold true.

figure b

To provide further illustration, (3B\(_{a}\)) with a preterite predicate cannot, and (3B\(_{b}\)) with a nonpast-nonperfective predicate can, be naturally followed by (4a). Both (3B\(_{a}\)) and (3B\(_{b}\)) can be naturally followed by (4b).

figure c
figure d

It can be said that, in Japanese, as long as negative clauses are concerned, the nonpast nonperfective is the default way to describe a situation in the past. The negative preterite, on the other hand, is subject to what can be called the “plausibility requirement”.

This work argues that the Japanese negative preterite predicate invariably expresses the existence (occurrence) of a “negative eventuality”, as opposed to the non-existence (non-occurrence) of eventualities, and that the plausibility requirement can be accounted for as a side effect of this feature. It will be furthermore argued that, while Japanese nonpast-tensed clauses generally specify that the topic time (in Klein’s 1994 sense) overlaps with or follows the topic time, this does not necessarily apply to nonpast-nonperfective clauses, making it possible for a negative nonpast-nonperfective clause to represent the non-existence of eventualities in a past topic time.

2 Existence of Negative Eventualities vs. Non-existence of (positive) Eventualities

It has been widely acknowledged in the literature that a negative clause may describe the existence (occurrence) of a negative eventuality, rather than the non-existence (non-occurrence) of eventualities (Krifka 1989; de Swart 1996; Przepiórkowski 1999; Bernard & Champollion 2018; Fábregas & González Rodríguez 2020; Higginbotham 2000; Zaradzki 2020). Among the most compelling pieces of evidence for “negative eventualities” are: (i) that a negative clause can be a complement of a perception verb like see, as in (5), and (ii) that a negative clause may occur in slots like “What happened is ...”, “... took place”, and “... is what they did”, as in (6).

figure e
figure f

The ontological nature of a “negative eventuality” has been a matter of extensive debate. This work adopts Bernard & Champollion’s (2018) idea that each set of eventualities P expressible with a clause nucleus has a negative counterpart, Neg(P), which contains all and only those eventualities which preclude—i.e., cannot co-exist in the same world with—every eventuality in P. When P is eventualities whereby Mary leaves, for example, Neg(P) is something like eventualities whereby Mary stays. Eventualities constituting P and Neg(P) will respectively be referred to as “P eventualities” and “anti-P eventualities”.

The incompatibility of an eventuality (a P eventuality) and its negative counterpart (an anti-P eventuality) may be accounted for with an axiom like (7), which mirrors the Law of Contradiction in classical logic.

figure g

Bernard & Champollion (2018) assign a meaning along the following lines to English not;Footnote 3 subscript E stands for “eventive”, and v is the type for eventualities.

figure h

Reference to anti-eventualities helps develop reasonable semantic representations for sentences like (5) and (6). It is an event of “anti-stopping” that is described as having been seen by a police officer, it is an event of “anti-visa-issuance” that is described as having happened, and so forth.

Now, if a negative clause may describe a negative eventuality, does it always do so? Does, say, the English adverb not always represent something like (8), or can it represent the classical Boolean negation, i.e. (9) where P stands for “propositional”, as well?

figure i

With Przepiórkowski (1999), Fábregas & González Rodríguez (2020), and Zaradzki (2020), I maintain that clausal negation may receive two distinct readings corresponding to propositional negation (= (9)) and eventive negation (= (8)). In a sentence like (10), the negation occurs in the complement of a perception verb and is forced to receive the eventive reading. In a sentence like (11), on the other hand, English not may, in theory, be either eventive or propositional.

figure j
figure k

A negative clause involving propositional negation can be said to express the non-occurrence of eventualities (NOE), and one involving eventive negation the occurrence of a negative eventuality (ONE). (11a) and (11b) are respectively paraphrases of the NOE and ONE readings of the sentence Mary did not dance.

I furthermore suggest that reference to a negative eventuality (corresponding to a dynamic event; see Sect. 4.3 for the case of stative eventualities)—i.e., the ONE reading of a (dynamic) negative clause—is highly constrained, and is available only when the occurrence of a corresponding positive eventuality is or was expected or at least plausible.

It has been commonly acknowledged that generally negative sentences are pragmatically more marked than their affirmative counterparts (Tian & Breheny 2019 and references therein). However, there seems to be a significant difference in the degree of markedness between sentences with regular (propositional) negation and ones with eventive negation. In a context where there has been no expectation for Mary to take a picture, let alone a picture of an eggplant, the negative sentence in (12a) would be a fairly strange thing to say. It nevertheless is judged as a true statement, if indeed Mary did not take a picture of an eggplant. The same goes with (12b), where the perception report as a whole is negated.

figure l

(13) situated in the same context, on the other hand, does not merely sound odder than (12a,b), but seems not to be true. It is not clear if it even counts as a false statement—it has a flavor of presupposition failure (see Miller 2003:297–299 and Zaradzki 2020:485 for relevant remarks).

figure m

(14) illustrates the (extra) markedness of eventive negation with a construction other than the direct perception report. Utterances like (14a) sound not only odd but also are perceived as non-true; (14b) sounds comparatively less odd and seems to count as a true statement.

figure n

(15), which has a structure parallel to that of (14), sounds rather natural, it being commonsensically plausible that Mary could have fulfilled the described action.

figure o

In sum, it seems fair to suppose that eventive negation is much more pragmatically constrained than that of propositional negation, and to posit the following generalization:

figure p

3 Proposal: The Japanese Preterite is Not Compatible with Propositional Negation

3.1 The Japanese Tense System

Japanese has a two-way distinction of tense: past and nonpast (also called present). The nonpast tense is marked with an inflectional ending: -(r)u for (affirmative) verbs and -i for adjectives, including negative predicates derived out of a verb with the suffix -(a)na (e.g. utawanai in (28a)). The past tense is marked with the marker -ta, which I take to be a particle following an infinitive predicate (Oshima 2014).

A nonpast-tensed dynamic predicate as a rule describes an event taking place after the relevant temporal anchoring point (typically the time of utterance), putting aside the habitual/generic interpretation. A nonpast-tensed stative predicate by default describes a state co-temporal with the anchoring point, but may also describe one that holds after it.

figure q

A past-tensed predicate, whether it is dynamic or stative, locates the described eventuality in the past relative to the anchoring point.

figure r

I assume that a tense poses a restriction on the topic time in Klein’s (1994) sense. (19) illustrates the case of the Japanese past tense marker -ta. TT and TU represent the topic time and the time of utterance, respectively. The logical predicate At is defined in (20) (cf. Condoravdi 2002:70). \(\tau \) represents the temporal trace function (Krifka 1989:97). \(\subseteq \) stands for the temporal inclusion. The material between braces represents non-proffered (not-at-issue) content.

figure s
figure t

I furthermore adopt the view that the Japanese nonpast tense does not code a temporal meaning, and it indicates “nonpastness” merely as an implicature arising from the absence of a past marker (cf. Sauerland 2002 on the English present tense). This supposition is not essential to the central claims of the present work, but it helps account for the distribution of the nonpast nonperfective to be discussed in Sect. 4 below.

figure u

One piece of evidence that the Japanese nonpast tense does not code temporal meaning is the observation that complex predicates carrying both a nonpast-tense feature and a past-tense feature, such as (22b,c), are interpreted as past-tensed, as if the nonpast-tense feature “gave way” to the past-tense feature.

figure v

The literal meanings of a past-tensed clause and a nonpast-tensed clause will look like (23) and (24).

figure w
figure x

The meanings of the rest of the constituents, and how they are combined with the meaning of a tense, are assumed to be as follows:

figure y
figure z

At the pragmatic level, the meaning in (24) is enriched into (27), where the implicated component is shaded.

figure aa

3.2 The Incompatibility of the Past Tense and the Propositional Negation

I propose that the negation in a Japanese negative preterite is invariably eventive, so that, for example, (28b) allows only the ONE reading while (28a) is ambiguous. The plausibility requirement for the negative preterite can be seen as an outcome of this feature (cf. 13)/(14)).

figure ab
figure ac
figure ad
figure ae

The lack of the NOE interpretation of the negative preterite likely has to do with the grammatical status/position of the past marker -ta. Historically, the marker -ta developed from the auxiliary tari, an archaic marker of perfect (Ogihara & Fukushima 2015). When a sentence with tari is negated, the negation occurs to its right, as in (32), where tar, the stem of tari, is followed by anu, a negative-attributive suffix (Kondo 2003).

figure af

The contemporary past marker -ta no longer retains its status as an inflecting word, and can only be preceded by negation. Some scholars, including Bloch (1946), Teramura (1984), and Tsujimura (2007), consider that -ta is—i.e. has grammaticalized into—an inflectional suffix directly following the predicate stem (the “attachment-to-stem” analysis). Others, including Shibatani (1990) and Shirota (1998), suppose that -ta is a particle or auxiliary that, like its predecessor tari, follows an infinitive form, an inflected form capable of heading a subordinate clause on its own, as in (33a,b) (the “attachment-to-infinitive” analysis).

figure ag

The infinitive form of a Type I verb (i.e. a verb whose stem ends with a consonant), such as au ‘see, meet’ (the stem = aw) and odoru ‘dance’ (the stem = odor), is formed by appending -i to the verb base (which may incur a phonotactically motivated sound change of the stem; e.g. aw \(+\) i \(\Rightarrow \) ai). I take -i here to an epenthetically inserted vowel, although it can alternatively be regarded as an inflectional suffix. The infinitive form of a Type II verb (i.e. a verb whose stem ends with a vowel), such as neru ‘sleep’ (the stem = ne) and hareru ‘clear up, get sunny’ (the stem = hare), is string-identical to the stem.

In Oshima (2014), I argued that -ta can be (though usually is not) separated from the verb to its left by an accent-phrase boundary, as in (34b), and argues that this lends support for the attachment-to-infinitive analysis; in (34a,b), braces indicate accent-phrase boundaries and downward arrows indicate accent falls.

figure ah

The verb to the left of -ta may be regarded either as the “host” or “complement” of -ta, depending on the premises regarding syntactic structure and headedness. I will regard it as a host for the sake of concreteness, but the choice here does not have direct bearings on the discussion.

The nonpast markers -(r)u and -i, as well as the negative-nonpast marker -en, on the other hand, can sensibly regarded as inflectional suffixes. (35) and (36) illustrate the compositions of the nonpast- and past-tensed plain (nonpolite) negative predicates whose base is aw ‘see, meet’ posited in Oshima (2014). Plus signs and slashes respectively indicate word-internal morpheme boundaries and word boundaries; “\(\Rightarrow \)” represents sound change, including the insertion of an epenthetic vowel, incurred by (morpho-)phonological rules. Subscript inf is meant to clarify the status of the expression as an inflected infinitive form.

figure ai
figure aj

The key point here is that -ta is separated from its host, which contains the negation, by a word boundary, while -i belongs to the same word as its host.

By and large the same goes with polite counterparts of nonpast- and past-tensed negative predicates. In nonpast ones, the tense feature occurs within the same word as the negation; in past ones, this is not the case.

figure ak
figure al
figure am

It seems quite plausible that the word boundary blocks negation in the host to take scope over -ta, thereby inducing the differing scopal behaviors of the nonpast and past tense markers.

figure an

Due to their semantic types, propositional negation (\(\langle t,t\rangle \)) must be applied after the closure of the eventuality variable, hence taking scope over the tense; eventive negation (\(\langle vt,vt\rangle \)), on the hand, may take scope under the tense (\(\langle vt,vt\rangle \)). The impossibility of the “Neg > Past (-ta)” pattern implies that the negation occurring in a preterite can only be eventive.

4 The Nonpast Nonperfective as an “Alternative Preterite”

The puzzle of the limited discourse-configurational distribution of the negative preterite has a flip side: the unexpectedly wide distribution of the negative nonpast nonperfective. I suggest that the Japanese nonpast nonperfective sometimes receives a “preterite-like” interpretation.

4.1 The -te iru form in its Perfect Use

The opposition between the so-called -te iru form (nonperfective form), and the simple form (perfective form) has been recognized to be central to the aspect system of Japanese. The -te iru form receives a wide array of interpretations, including (i) resulting state (also called resultative perfect), (ii) progressive, and (iii) habitual (e.g. Kudo 2020). Among the various uses of -te iru, the one that most directly concerns the purposes of the current work is the one labeled “perfect” in such works as Shirai (2000) and Kudo (2020) (alternative labels for this use include “existential perfect”, “experience (keiken)”, and “retrospection (kaisoo)”).

Providing examples like (41a–c), Teramura (1984:131) maintains that the function of -te iru in its perfect use is to describe “an event in the past that has significance on the present time (genzai ni igi o motsu kako no jishoo)”.

figure ao

A nonpast-tensed -te iru form in its perfect use has a meaning rather similar to that of the corresponding preterite, much like how an English present-perfect clause is similar in meaning to its preterite counterpart (e.g., Ken has read the book vs. Ken read the book). Given this, it is tempting to suppose that a -te iru form occurring in a direct answer to a past-tensed interrogative, such as the instance in (1B\(_b\)), receives the perfect interpretation (I will dismiss this view below, however).

I assume that the -te iru perfect is by and large synonymous to the English have -ed perfect. Here I adopt Parsons’s (1990) resultativity-based analysis of the perfect aspect, according to which it describes the resultant state of an eventuality, i.e., an abstract state whereby some eventuality’s “having occurred”. The meaning of i (the stem of iru) in its perfect use is taken to be something like (42); RS stands for “resultant state”.

figure ap

Nonpast/past-tensed perfect clauses (43a,b) will be taken to have the meanings in (44a,b) respectively, with the derivational process schematized in (45) (the gerund marker -te is considered to be semantically vacuous).

figure aq
figure ar
figure as

4.2 The Nonpast Nonperfective as an “Alternative Preterite”

Some instances of nonpast-nonperfective predicates appear to receive a “past-like” interpretation that is to be distinguished from the perfect interpretation.

Under the assumption that the nonperfective auxiliary in (46B\(_{b}\)) indicates the perfect aspect, the meanings of the boldfaced parts of (46B\(_{a,b}\)) should look like (47a,b).

figure at
figure au

It is implausible, however, that the relevant parts of (46B\(_{a}\)) and (46B\(_{b}\)) are construed as being “about” different temporal scenes, the topic time set in the past (relative to the utterance time) and in the nonpast respectively. Affirmative response (46B\(_{a}\)) to question (46A) cannot be naturally replaced with its nonpast-nonperfective variant, i.e. (46B\(_{a\text {'}}\)), suggesting that the topic time has to be set in the past in this context. There is no evident reason that the same does not happen when a negative response is made to the same question.

It is noteworthy that some instances of affirmative nonpast-nonperfective clauses, too, seem to make reference to a past topic time. Such instances are commonly found in written historical and biographical accounts, as exemplified with (48), a part of a Wikipedia article on Mahatma GandhiFootnote 4 (see also (41c) from Teramura 1984).

figure av

I suggest that, on top of its perfect(, resultative, progressive, \(\ldots \)) use(s), iru has a “quasi-past” use, which specifies that the topic time is in the past.

figure aw

The topic-time restriction posed by iru in its quasi-past use conflicts with, and hence suppresses, the “nonpastness” implicature that a nonpast predicate usually induces.

To illustrate, (50a) and (50b) each have two possible logical translations (putting aside the ONE interpretations of (50b)): (51a,b) for (50a) and (52a,b) for (50b). (51b) and (52b) can be characterized as the “alternative preterite” reading. Note that “i” and “-i” respectively refer to the nonperfective auxiliary stem and the nonpast tense suffix following an adjectival stem.

figure ax
figure ay
figure az

On the (a) interpretation, (50a,b) are “about” a temporal scene where the resultant state of an event whereby Ken sees Mari, described by the pre-tensed clause, held or did not hold. On the (b) interpretation—i.e. the alternative-preterite interpretation—(50a,b) are “about” the temporal scene where an event whereby Ken sees Mari occurred or did not occur. I take (46B\(_{b}{} \)), as well as the second sentence in (48), to receive the alternative-preterite interpretation.Footnote 5

There is additional language-internal evidence that iru may indicate pastness: it serves to indicate temporal anteriority in some types of conditional clauses in which a tensed clause cannot occur. Japanese has several markers of conditional clauses, including (r)eba, tara, and naraba. (R)eba and tara respectively follow a verbal base and a verb in its infinitive form (Oshima 2014), and thus neither can be combined with a tensed clause; naraba, on the other hand, follows a tensed clause.

figure bb

In conditional constructions with (r)eba and tara, if the antecedent describes a possibly true situation in the past (relative to the utterance time), the predicate must involve iru.Footnote 6

figure bd
figure be

(55b) is more naturally paraphrased with (56a) with a preterite naraba-conditional clause than with (56b) with a nonpast-nonperfective one.

figure bf

It is thus natural to suppose that iru in the antecedent of (55b) indicates pastness in the way the past marker ta does in environments where it can occur, such as the antecedent of (56a).

It is worth noting that a rather similar phenomenon is seen in English: the have -ed construction, which typically expresses the perfect aspect, can be deployed to express mere temporal anteriority in environments where finite past forms cannot occur, as in (57b):

figure bg

I suggest that, the -ta preterite being the default/unmarked means of describing a situation in the past, the quasi-past meaning of i is mobilized only for special purposes. One is to compensate for the inability of a negative preterite to describe the non-occurrence of eventualities. Another, typically seen in formal writings, is to signal a marked discourse relation (rhetorical relation), such as exemplification or supplementation, between the clause and the surrounding discourse segments, as in (48). Yet another is to form a (r)eba or tara-conditional clause describing a past situation.

4.3 Negative Preterites with a Stative Base

When the base of a predicate is stative (a stative verb, an adjective, or a nominal predicate), the addition of -te iru is blocked, or has no or only a subtle semantic effect.

figure bh

A negative preterite form with a stative base does not implicate the plausibility of the logical contradiction of the propositional content in the past; the naturalness of (59B)/(60B) illustrates this point.

figure bi
figure bj

It is not clear to me if this implies that a negative preterite with a stative base allows the NOE interpretation. It seems plausible that for a negative predicate with a stative base, the ONE reading is not marked, or even is preferred to the NOE reading, and thus is not pragmatically constrained in the same way as that of a negative predicate with a dynamic base is. This supposition is motivated by the observation that “anti-states” can often be lexically coded, unlike “anti-(dynamic-)events”. In the case of English, for an adjective expressing stative concept S, it tends to be possible to derive, with prefixes like non- and in-, another adjective expressing “anti-S” (e.g., non-American, inappropriate). The same does not go with verbs, which generally express a dynamic event.

Consequently, the lack of the plausibility implication in an utterance like (60B) does not necessary undermine the generalization that -ta cannot be outscoped by (propositional) negation to its left.

5 Conclusion

It was argued that in Japanese, as long as situations in the past are concerned, “non-occurrence of eventualities (NOE)” and “occurrence of a negative eventuality (ONE)” are coded differently. NOE is invariably coded with a nonpast-nonperfective form, and ONE is typically coded with a preteritie (past-perfective) form. It was also proposed that a nonpast-nonperfective form may indicate that the topic time is in the past, thereby inducing an “alternative-preterite” interpretation.

While various pieces of evidence have been put forth in the literature for the existence of negative eventualities as linguistically expressible objects, explicit markers or constructions favoring one of the NOE and ONE interpretations and deterring the other have hardly been discussed. The analysis presented in this work suggests that examination of Japanese data, and search for phenomena comparable to the Japanese regular-preterite/alternative-preterite opposition in other languages, have good potential to deepen our understanding of “negative eventualities”.