Skip to main content

Reported Speech and Tense in Japanese: Why There Is No Backshift

Japanese reported speech does not backshift tense: the quoted clause keeps the tense the original speaker used. It is anchored to that speaker's moment, not yours.1 English speakers expect "he said he was going" to push the present into the past, but Japanese leaves the inner verb exactly as it was first spoken.

Overview

This article assumes you already know how to quote with と and って. The foundation article on Japanese quotation with と covers that. The question here is narrower and trips up almost everyone arriving from English: once you put a quote under a past reporting verb, does the tense inside the quote change? In Japanese, it does not.

This is an N3-level concern. Not because the quoting form is advanced, since it is introduced much earlier, but because the tense behaviour is the part learners get wrong. The phenomenon sits inside the broader picture of how Japanese marks tense and aspect.

The English backshift habit you must unlearn

English has a rule grammarians call the sequence of tenses (SOT), or consecutio temporum (Latin for "sequence of times"). Under this rule, a present-tense complement under a past main verb is normally converted to past. Ogihara describes the pattern as "a situation where a past tense occurs immediately under another past tense, but the lower past tense is interpreted as referring to a time simultaneous with the time referred to by the higher past tense."1

Take his own example: John said that Mary was sick. The default reading, which Ogihara calls the simultaneous interpretation, is "that the time at which Mary is allegedly sick is simultaneous with the time of John's utterance."1 English uses a past form (was) to describe a moment that is simultaneous with said, not earlier than it.

That conversion is the habit to unlearn. The Japanese term for SOT is 時制の一致 (jisei no itchi, "tense agreement"). The claim of this article is that Japanese has 時制の一致がない: no tense agreement.

What "no backshift" means in one sentence

The tense morpheme, or tense marker, inside the quoted clause is fixed to the moment of the original utterance or thought. The past tense of the reporting verb does not reach inside to change it.1

English and Japanese arrive at the same meaning by opposite routes. On Ogihara's analysis, English deletes the embedded past under SOT, while a Japanese non-past embedded clause is read relative to the matrix event; the surface tense differs because the mechanisms differ.1

The rule: the quoted clause keeps its original tense

The foundation article on Japanese quotation with と covers how to attach と and って, and how to mark who said what. This section covers only one thing: what the tense inside the quote does once the quote is closed by a past reporting verb. The answer is that it does nothing. It stays as first spoken.

Non-past inside a past report

When the original speaker used a non-past verb, that verb stays non-past even under a past reporting verb such as 言っていました.

病院びょういんくとっていました。2
"Yui said she's going to the hospital."

The embedded verb 行く ("go") is non-past and unchanged, even though the reporting verb 言っていました ("was saying/said") is past. English cannot leave it untouched; it must produce "she's going" or "she would go." Japanese does not.

九時くじいえかえるとっていました。2
"Fred said he will go home at 9."

Non-past 帰る ("go home/return") is preserved. The source even renders the English with the future "will go home," the exact form an English speaker is tempted to backshift into "would go home."

Adjectives behave identically. A non-past adjective stays unchanged inside a quote closed by past 言った ("said").

さむい」とアリスが田中たなかった。3
"'Cold,' Alice said to Tanaka."

Past inside a past report

A past tense inside the quote is not a "double past." It means the inner event was already past at the original speaker's moment. The embedded past is read relative to the reporting event. It places the inner event before that report.1

Ogihara's minimal academic pair shows this directly. His romanization uses doubled vowels rather than Modified Hepburn. It is preserved verbatim here as the cited form, with the modern kana alongside.

Jon-wa [Mearii-ga byooki-datta] to it-ta. (modern kana: ジョンは「メアリーが病気だった」と言った)1
"John said that Mary had been sick."

The embedded predicate is past (byooki-datta / 病気だった, "was sick"), and the reading is the shifted one only: Mary's sickness is located before John's saying. English reaches the same meaning with the pluperfect "had been." Japanese did not convert anything. It simply placed the inner event earlier than the matrix event by using past tense.

The same structure appears in learner-facing material.

太郎たろう花子はなこ病気びょうきだったとった。4
"Tarou said that Hanako had been sick."

The past だった ("was") forces the shifted interpretation. The source spells this out as implying that Hanako was no longer sick when Tarou spoke.4 The tense mechanics of this past-inside-a-quote draw on the plain past た-form.

The contrast table: English vs Japanese

Ogihara gives the cleanest contrast in the literature, reproduced here with his verbatim forms. English backshifts the same content to past. Japanese leaves it present.

FormReading
EnglishJohn said that Mary was sick.Simultaneous interpretation: Mary sick at the time John spoke.1
JapaneseJon-wa [Mearii-ga byooki-da] to it-ta. (modern kana: ジョンは「メアリーが病気だ」と言った)"John TOP Mary NOM be-sick PRES that say PAST"; simultaneous interpretation only.1

The key observation, in Ogihara's words, is that this Japanese sentence "has a present tense morpheme in the verb complement clause."1 English backshifted the same content to was; Japanese left it present as byooki-da / 病気だ.

A learner-facing source gives the non-past member of the pair as well: 太郎は花子が病気だと言った, glossed "Tarou said that Hanako was sick." In that sentence, the present だ "expresses that Hanako was sick when Tarou spoke," the simultaneous reading.4

For a verbal rather than adjectival comparison, set the non-past 行く and 帰る examples above against the English "would go" and "would go home" that a backshifting translator produces. The Japanese forms are sourced; the backshifted English is the well-known SOT output Ogihara describes,1 not a separate Japanese sentence.

The backshifted English is a translation, not a model

When a source renders 帰ると言っていました as "Fred said he will go home," the smooth English may use "would." Read that "would" as English grammar applying its own SOT rule. Do not let it pull you into changing the Japanese 帰る; the Japanese stays non-past no matter how the English column is phrased.2

Nuance and usage contexts

Anchoring: whose "now" is the quoted tense measured from

The embedded tense is read from the matrix event's clock, meaning the original speaker's moment, not the present reporter's moment. Here is the relative-tense rule: a past tense in the subordinate clause puts its event before the matrix event. A non-past tense puts it at the same time as the matrix event, or in the future relative to it.4

This is why swapping present and past changes the temporal location. The non-past byooki-da / 病気だ forces the simultaneous reading, and the past byooki-datta / 病気だった forces the shifted, prior reading, under the same matrix it-ta.1

The non-past examples above show the anchoring at work. 行く and 帰る are read as simultaneous with, or later than, the saying. So "she's going" and "he will go home" both measure from the speaker's moment, not from when you relay the report. The tense is the anchor. You do not re-date it to your own present.

The timeline makes the contrast with English visible.

〜と言った vs 〜と言っていた

This freezing applies only to the inner clause. The reporting verb on the outside is fully live and shifts like any other verb. The choice between 言った and 言っていた is made entirely on that outer verb. It does not touch the frozen inner clause.2

The continuous-past 言っていた / 言っていました carries a nuance of "relaying what I heard earlier." Gokigen draws the contrast between the two outer forms directly. 言っていました "often carries a reported-speech nuance," telling the listener "what someone said earlier, with a sense of relaying information." By contrast, 言いました "is a simple past statement meaning 'said,' without that reported-information feel."2

The casual reduction of 言っていた / 言っていました is 言ってた (itteta).2

鈴木すずきさん、お寿司すしきじゃないとっていました。5
"Suzuki-san said she doesn't like sushi."

Whatever you do to the outer verb, inner 好きじゃない ("doesn't like") stays non-past. Only the outside moved into the continuous past; the inside is frozen.

Same principle in other embedded clauses

The no-backshift rule is one example of a general property: Japanese embedded clauses carry relative tense, read against the matrix event. The same independence shows up beyond verb-complement clauses, including relative clauses.

The Keio Snippet gives a relative-clause demonstration that embedded tense is relative, not absolute. It says an embedded tense is ordered against the matrix event, "the event time of relative clause … in relation to the event time of the matrix clause."6

The Keio Snippet's headline is an exception, used here only for the general point

The Snippet's specific finding is a narrow exception: a past-progressive embedded clause under a non-past matrix forces an utterance-relative reading.6 This article cites it only to establish the general fact that embedded tense in Japanese is relative. Do not read the exception back as the everyday rule; the everyday rule is the relative-tense, no-backshift behaviour shown throughout this article.

Good to know

Mnemonic: tense reports the speaker's clock, not yours

Treat the quote as a photograph: you cannot re-date it. The tense inside is fixed at the original utterance moment. Your job in reporting it is to leave it alone. Mechanically, the embedded clause is read relative to the matrix (saying) event. A non-past inner clause sits at the same time as or after that event, and a past inner clause sits before it.41

Backshifting the inner verb to agree with past 言った

The most common error is converting the inner verb to past so it "agrees" with the past reporting verb, as it would in English. This is not a stylistic agreement change; it relocates the event in time and changes the meaning.

Writing 行った ("went") instead of 行く makes the sentence say that Yui said she went to the hospital, because an embedded past forces the prior-to-saying reading.1 To keep the meaning "Yui said she's going to the hospital," the inner verb must stay non-past.

病院びょういんくとっていました。2
"Yui said she's going to the hospital."

How this interacts with hearsay 〜そうだ and quoted thought 〜と思った

The frozen-inner-clause logic also applies to other reporting frames. Quoted thought with と思った ("thought that") and "heard that" with 〜と聞いた ("heard that") anchor the inner tense to the original mental or utterance event in exactly the same way. The source shows a non-past inner clause under past 聞いた ("heard"):

今日きょう授業じゅぎょうがない」と先生せんせいからいたんだけど。3
"It is that I heard from the teacher, 'There is no class today.'"

Inner ない ("does not exist/is not") is non-past under past 聞いた. The hearsay form 〜そうだ follows the same no-backshift logic. See the dedicated treatment of that form for its mechanics, which are not re-taught here.

See also

References

Footnotes

  1. Ogihara, Toshiyuki. "The Semantics of Tense in Embedded Clauses." University of Washington (author's posted manuscript; the analysis also appears in Ogihara, Tense, Attitudes, and Scope, Kluwer Academic Publishers, 1996). https://faculty.washington.edu/ogihara/papers/Ogihara_tense.pdf 2 3 4 5 6 7 8 9 10 11 12 13 14

  2. "How to Say 'Someone Said That…' in Japanese | Using to itteimashita (〜と言っていました)." Gokigen Japanese (blog.gokigen.jp). https://blog.gokigen.jp/how-to-say-someone-said-that-in-japanese-using-to-itteimashita/ 2 3 4 5 6 7

  3. Kim, Tae. "Making and performing actions on quotations." Tae Kim's Guide to Learning Japanese. https://www.guidetojapanese.org/quotation.html 2

  4. "Tenses." Japanese with Anime. https://www.japanesewithanime.com/2020/08/tenses.html 2 3 4 5

  5. "How to Say 'Someone said~' in Japanese - JLPT N4 Grammar: と言っていました." Coto Academy. https://cotoacademy.com/say-someone-said-japanese-jlpt-n4-grammar/

  6. Kawahara, Shigeto, et al. "A gap in the interpretation of embedded tense in Japanese." Snippet (Keio University posting). http://user.keio.ac.jp/~kawahara/pdf/snippet.pdf 2