The other day Matt posted on the Classical particles traditionally described as “past” (though many seem to be about aspect rather than time): namely ki, keri, nu, tsu, tari, ri. I commented on some unusual points of Fujii Sadakazu’s analysis, which I’m more than a little skeptical about. Then it struck me that I should take a closer look at what a linguist has to say; so I read more carefully the relevant sections of Frellesvig’s A History of the Japanese Language, and summarized them below. For contrastive illustration, I’ve used my photocopy of a grammar table from some handbook or another as a typical representative of traditional (school) grammar.
たり = -(i)tar-i “stative” or “resulting past”
Traditional grammar: sonzoku “continuing state” or kanryō “completed” (= modern -ta “past, resulting past”, -te shimatta “finished past”). Follows ren’yōkei.
Frellesvig: first “stative”; later, also “resulting past” (18.104.22.168.3, 8.4.2).
A fusion of Old Japanese (OJ) -te ar-i. It’s originally a “stative”, very much like the modern -te ar-u: both the fused and the non-fused forms describe a situation as a state. This often (but not always) denotes the state resulting from a (past) action, and thus the stative has been called a “resultative”, “durative”, “progressive” etc. Furthermore, the “resulting state” nuance implies a completed action, and therefore this is has also been called “perfect (past)“.
(Aside: “perfect” and “perfective” are exceedingly confusing grammatical terms, very similar but annoyingly distinct. The “perfect” we just mentioned above has various meanings, but usually it denotes “the continuing present relevance of a past situation” – the action was completed in the past, but we’re talking about the present consequences. As for “perfective”, it’s used to talk about any completed event as a singular point in time – an action as a whole, be it past, present, or future. I’ll avoid both terms, changing Frellesvig’s “perfective” (see below) to “complete”, and “perfect (past)” to the awkward but less misleading “resulting past”. The traditional-grammar term kanryō seems to cover both meanings.)
So there’s two meanings, stative and resulting past. One important point is that the latter arose gradually. About the Heian period (Early Middle Japanese, EMJ) the fused form (i)tar-i started being used for tense (past vs. present), rather than just aspectual meaning (long, continuing state vs. point in time). By Muromachi (Late Middle Japanese, LMJ) this shift had already produced the modern suffix for past, -(i)ta, replacing in speech the plain past -(i)ki and the modal past -(i)ker-i.
り = -er-i “stative”
Traditional: Exactly the same as たり: sonzoku “continuing state” or kanryō “completed”. Follows izenkei.
Frellesvig: This is also an stative, older than -(i)tar-i. It derives from OJ -yer-i, which is itself derived from a construction combining the verb ar-i (just like -(i)tar-i). From EMJ onwards, -er-i is gradually replaced by -(i)tar-i; but they coexist for a long time. -(i)tar-i can be used with all verbs, but -er-i only with consonant roots.
Unlike the traditional grammar, Frellesvig says -er-i remained an atemporal stative, and never acquired the “resulting past” meaning that -(i)tar-i eventually got; -er-i was always “atemporal” (=purely aspectual).
Is there any difference in nuance in choosing -(i)tar-i or -er-i to mark the stative of a (consonant-root) verb? If so it’s still unclear, and a point of contention.
つ = -(i)te: complete; starting; assertion; or sequenced narration (also: -(i)tar-ite: recent past)
Traditional: Completed (kanryō), emphasis (kyōi), or parallel sequence (heiretsu).
Frellesvig (22.214.171.124, 8.4.2): There are two uses. One is to consider an action as a whole, from beginning to end. This may be looked “backwards”, as a completed action (mi-tu “has seen”); or “forwards”, as something about to start (inceptive: naki-tu “begins to sing”). (Both of these are common with the “perfective aspect” in other languages.)
The other use is as an assertion or affirmation. This is the original meaning, since the morpheme derives from the copula to. Because of this, -(i)te never combines with negative suffixes; but often combines with the conjectural, as -(i)te-m-u.
The traditional notion of heiretsu is acknowledge by Frellesvig as Takeuchi’s “sequenced narration”.
Recall that -(i)tar-i was originally aspectual (a stative), but latter acquired a tense meaning (resulting past). -(i)te likewise got a tense meaning, specifically when combined with -(i)tar- to make -(i)tar-ite- “recent past”.
ぬ = -(i)n-u: complete; starting; assertion; or sequenced narration (also: (i)n-itar-i: limited control)
Traditional: Exactly the same as つ: Completed (kanryō), emphasis (kyōi), or parallel sequence (heiretsu).
Frellesvig (126.96.36.199, 8.4.2): The same as -(i)te above. Whereas the one came from copula to, the other came from copula ni.
- Completed, past: tir-in-u “has fallen”
- Completed, future (inceptive): nak-in-u “begins to sing”
- Assertion: mit-ik-in-amu “it will surely rise”
As shown, when combined with conjectural -(a)mu, it becomes the familiar -(i)n-amu.
It’s a point of debate whether there’s a difference in meaning between -(i)n-u and -(i)te. They may be in complementary distribution (same meaning, different contexts of use): Frellesvig cites John Whitman’s observation that -(i)te is used for transitive and intransitive unergative verbs, while -(i)n-u goes with intransitive unaccusative ones. There is a small number of verbs (11 in total, listed on p. 68) which break the rule and may occur with either suffix, perhaps due to semantic motivations.
However, one clear distinction in meaning arose in post-Heian Japanese, when combined with the new temporal nuance of (i)tar-i. As we have seen, (i)te combined after it to build (i)tar-ite “recent past”. By contrast, (i)n-u combined before it to make (i)n-itar-i, which denotes “limited control”, i.e. difficulty or unintentionality.
き = -(i)ki: simple past
Traditional: Past (kako). Follows ren’yōkei.
Frellesvig: Simple past. It’s the past. Something occurs before a reference time point, often the moment of speaking.
The only complicated thing about this suffix is its irregular inflection (on the traditional system: se, -, ki, si, sika, –). Very early poetry (in the Kojiki and Nihon Shoki) preserved archaic forms -kyeku and -kyeba, which soon disappeared. The forms -siku and -seba survived until EMJ.
It can combine with the conjectural -(a)m-u to make past conjectural -(i)kem-u.
けり = -(i)ker-i: speaker commitment, resulting past
Traditional: Past (kako) or exclamation (eitan) or both. Also past hearsay (modern ~ta sou da).
Frellesvig: It’s complicated. (188.8.131.52.2, 184.108.40.206)
Deriving from OJ -(i)kyer-i, it never combines with the simple past -(i)ki, being in clear opposition to it. It also never combines with the conjectural -(a)mu (unlike -(i)ki → -(i)kem-u).
Everyone agrees that this form combines past tense with some sort of subjective mood, though the exact nuance of this modality has been notoriously hard to pinpoint, and there are a lot of conflicting proposals. Frellesvig reject the “hearsay” theory of the modal past (and the corresponding “direct experience” nuance for simple past), for lack of concrete evidence. Rather, what he identifies as the common point of historical -(i)ker-i usage is a mood of speaker commitment—a feeling of “I’m telling you, it was so”. At times this mood may be such a focus that -(i)ker-i isn’t even used for past, only for commitment (which matches the traditional analysis as “emphasis”).
One point in favor of this theory is that the modal past is never used with the conditional (unlike the simple past), which is what we’d expect for a subjective modality. Another one is that, in old texts, simple past is used in imperial edicts (senmyō) as well as prayers (norito), whereas the modal past is used only in the former. This is because the edicts are “addresser focused” (from the emperor), while prayers are “addressee focused” (to the spirits). Furthermore, in edicts, the modal past is used in speech with speaker commitment (what I did, what I lived), while the simple past is used to frame non-speaker-oriented events in the distant past (often ancient feats of Gods or past Emperors).
The modal past has further a separate meaning as yet another marker of resulting past (“perfect”), which Frellesvig analyzes as a different construction, which ended up being homophonic: -ko “come to” + -yer-i stative > -(i)kyer-i > -(i)ker-i “has come to”.
Frellesvig organizes the original OJ verbal suffixes along two dimensions or systems, one for aspect (including negation), and the other for tense+mood. These determine what kind of suffix may combine with others; once a morpheme is chosen, the others of the same dimension can’t occur in the same combination. Here’s tabular reproduction of his diagram:
Items stacked vertically may not combine: so the complete (“perfective”) suffixes, -i(te) and -(i)n-u, never combine with a negative like -(a)zu, or a stative (-(i)tar-i). The simple past -(i)ki does combine with the conjectural -(a)mu, but the modal past -(i)ker-i never combines with either, nor does the subjunctive (-(a)masi).