Ellipsis and Implicit Reference in Japanese
Ellipsis and implicit reference in Japanese means systematically dropping any element a listener can recover from context. A "complete" sentence may have no stated subject, no object, and sometimes barely a verb, yet still be perfectly clear to a native speaker. Japanese is a high-context language: it leaves unsaid whatever the listener already knows. Over-stating recoverable information is not just redundant. It can sound condescending.12
Subject-dropping is the most-studied part of this system and has its own deep dive. This article assumes you have that background and maps the rest: the implied object and verb, topic-chain referents, and whole-clause omissions that make Japanese feel so sparse.3
What Japanese can leave unsaid
The Japanese metalinguistic term for ellipsis is 省略 (shōryaku), literally "omission" or "abbreviation." Reference grammars use 省略 for the dropping of any recoverable element, and 主語省略 (shugo-shōryaku) for the subject-omission subcase.4 In linguistics, the unpronounced argument is analyzed as a null pronoun or zero pronoun (ゼロ代名詞). The phenomenon is called null anaphora and is treated as a core syntactic property of the language.5
Omission is not free deletion. The constraint is recoverability: omitted information must be retrievable. As one reference puts it, "if omitted information cannot be retrieved and correctly interpreted by a speaker, said omission is consequentially ungrammatical."4 An omission analysis therefore treats the dropped element as present in the underlying structure, not absent.4
A corpus analysis of a BCCWJ-derived predicate-argument dataset by Ishizuki and colleagues (2024) "revealed that 37% of arguments, such as subjects and objects in Japanese, are omitted."3 In other words, ellipsis is a system the intermediate learner has to master, not an edge case to skip.
The omission inventory
Japanese can leave more than the subject empty. The argument-ellipsis study states the scope directly: "In Japanese, every argument of a predicate can potentially be omitted depending on the context."3 Its corpus annotation targets the nominative (subject), accusative (direct object), and dative (indirect object) arguments. All three are attested as omitted in a balanced corpus.3
Beyond the arguments, the predicate itself can be elided. A bare-noun reply leaves the verb or copula unexpressed but recoverable. This case is taken up below in "Predicate and verb ellipsis."64
A no-context baseline keeps every slot filled:
彼はもう彼のりんごを食べた。3
"He already ate his apple."
Give the same content a supporting context, and both the subject and the object drop. The prior sentence has already supplied them:
彼のりんごはない。もう食べたので。3
"He doesn't have his apple. (It's) because (he) already ate (it)."
In that second sentence, both the subject (彼は) and the object (彼のりんごを) are gone. They are recovered from the first sentence, and the study glosses the omitted material in parentheses.3
Why subject-drop lives elsewhere
Subject ellipsis, also called pro-drop or 主語省略, is the textbook entry point. The reference notes that "one of the most common omissions in Japanese is removing the subject, especially when indicating the first person, since subjects are usually implied in the context of the conversation."2 It is the single most common omission, and it has its own dedicated treatment.
This article takes subject-drop as given. The mechanics of how a dropped subject is recovered, and when it is safe to drop, belong to the dedicated Dropped Subjects in Japanese: Pro-Drop Explained article. Here the focus is the remaining slots: the object, the indirect object, the predicate, and the topic-chain referents that span multiple clauses.
Object and indirect-object ellipsis
Dropping the を-object
The accusative slot is one of the three argument positions the corpus study targets. It is shown dropping under context: when the object 彼のりんごを ("his apple") is recoverable from prior discourse, it simply vanishes.3 A chunk-level rule governs the drop: "when the phrase ringo-wo (apple) is dropped... its modifier kare-no (his) should also be omitted."3 The object and any material modifying it leave together.
彼のりんごを食べた。 → 食べた。3
"(He) ate (his apple)." → with both arguments recoverable, the surface form is just 食べた.
Where Japanese drops the object outright, English still requires an overt object pronoun, as in "I ate it." The English slot cannot go empty (see "English contrast").7
The same recovery happens across a question and its answer. The exchange below is a constructed illustration of the pattern. The bare-verb form 食べた and its parenthesized-argument analysis are sourced from the corpus study:3
A: 食べた？ B: うん、食べた。3
"A: Did (you) eat (it)? B: Yeah, (I) ate (it)." (A's turn is a constructed question; the bare-verb answer matches the sourced form.)
Indirect objects and obliques
The dative slot, the に or へ target, is the third argument position the corpus study annotates as omittable. It follows the same recoverability principle as the subject and object.3 The giving and receiving verbs give a concrete case. With くれる, the recipient defaults to the speaker or the speaker's in-group. That is why the 私に ("to me") indirect object is routinely dropped: "if the giver or the receiver is you yourself, you can often omit the 私に part in Japanese."8
ジェニーが（私に）ガムをくれる。8
"Jenny gives (me) gum."
The recipient 私に is shown in parentheses because くれる already points the action inward toward the speaker, so the dative drops without loss.8
Predicate and verb ellipsis
Bare-noun replies and 私も
The predicate can be fully recoverable from the prior turn, leaving only the new or contrasted element pronounced. The clearest case is the unagi-sentence (鰻文 / うなぎ文), the canonical example of a noun-only utterance standing in for a full predicate. According to Obana (2001, p. 726), as reported in a comparative study of the construction, the term 鰻文 (unagi-bun) "was coined by Okutsu (1978)."6910
僕はうなぎだ。611
"As for me, eel." (lit. "I am an eel"; in a restaurant, "I'll have the eel.")
An unagi-sentence "has the same structure as a typical copular sentence, namely [NP1 wa NP2 da]." However, "NP2 cannot directly be explained as being descriptive or identifying of NP1. Instead, the relation between NP1 and NP2 is context dependent."6
Read without context, the sentence is strange, since the typical copula reading "would imply that the speaker identifies as an eel."6 Only in context does it resolve. In a restaurant deciding what to order, it means "I will order the eel"; among people discussing disliked foods, "I dislike eel"; among anglers planning a trip, "I will catch eel."6 The construction has been treated in the literature as elliptic, with "transformational approaches, using some kind of ellipsis," reconstructing the missing predicate.6
The pattern underlying these bare-noun replies is copula ellipsis. In other words, だ is omitted from a nominal predicate. This is allowed when "the sentence is simply a statement of 'X is Y'" with no added modality and the predicate "is readily retrievable."4
ううん、あれは猿よ。4
"No, that there is a monkey."
The bare noun 猿 carries the whole predicate when the "X is Y" reading is fully recoverable. That recoverability is what licenses the drop.4
セス先生が飼っているのは猫。4
"The animal Seth-sensei keeps is a cat."
Here the predicate-final だ is omitted, and the bare noun 猫 stands in for it.4 The same logic extends to additive replies like 私も ("me too"), built on the も particle. In those replies, the predicate from the prior turn is inherited wholesale, and only the new participant is voiced.
Trailing-off and clause-final omission
The same recoverability logic lets a speaker stop before the main predicate. A clause can trail off at a connective such as が, けど, or ので with the verb unsaid. The argument-ellipsis study's own context example ends on a 〜ので clause whose main predicate is left for the listener to supply.3
もう食べたので。3
"(It's) because (he) already ate (it)."
The main clause this explains, that he has no apple, is recoverable from the prior sentence and is left unsaid.3 Stopping before the conclusion is a softening device. Leaving the conclusion unstated hands it to the listener, a point the discourse section returns to.
Recovering what was dropped: zero anaphora
Topic chains: は sets it, the rest inherits it
The recovery mechanism is null anaphora: a zero pronoun in the sentence takes its referent from the discourse. The syntactic treatment of how that null element is represented and interpreted is a core part of Japanese grammar.5 The topic marker は is the engine of multi-clause omission. Because は marks a topic rather than a grammatical subject, once a referent is established with は, subsequent clauses inherit it as the default referent without re-marking. Overt re-statement of an already-established subject is marked and "sounds unnatural," which is the flip side of saying the chain carries the topic silently.2
A single topic, set once, threads through every following clause until something new is marked. The diagram below traces that inheritance.
彼のりんごはない。もう食べたので。3
"He doesn't have his apple. (It's) because (he) already ate (it)."
The first clause's referents carry into the second. 彼は and 彼のりんごを are both inherited and therefore unexpressed in 食べた.3
Discourse, situation, and shared knowledge
Recovery draws on more than the prior sentence. The licensing condition is that the dropped element be "implied in the context of the conversation" and "clear in context." That covers both prior discourse and the shared situation.2 Honorific and directional morphology act as recovery cues that pin the dropped referent.
Giving and receiving direction is one cue. くれる points the action inward toward the speaker, so a くれる sentence with no stated recipient is read as "...gives me...," letting 私に drop.8
The desire form 〜たい is another cue. It is restricted to the speaker: "you can only use the 「たい」 form for the first-person because you cannot read other people's mind to see what they want to do."12 A bare 食べたい is therefore recovered as "I want to eat," with no subject needed.1312
クレープが食べたいから。13
"Because (I) want to eat crepe."
The 〜たい form binds the desire to the speaker, so the dropped subject is recovered as first person without any pronoun.1312
When omission becomes genuinely ambiguous
Ellipsis has limits. When the antecedent is lost, or two referents compete, the zero pronoun cannot be resolved. The omission then becomes unclear. As a general principle, if omitted information "cannot be retrieved and correctly interpreted," the omission "is consequentially ungrammatical."4
The repair is to re-supply just enough to disambiguate: re-state the topic with は, or use a demonstrative such as それ or あれ to re-anchor the referent.4 That is exactly the material a copula-ellipsis or bare-reply utterance otherwise leaves out. The speaker can restore it the moment context stops carrying the load.
The high-context logic: ellipsis as politeness
Saying less respects the listener
The pragmatic core is that stating what the listener already knows implies they could not infer it. Explicitly stating a known subject "sounds unnatural" because it implies contrast. This is the same は-versus-が contrast logic that governs which marker an overt subject would take. When no contrast is intended, omission yields "the most natural sentence."2
Over-specifying the second person is the sharpest case. Naming the listener with あなた when their identity is obvious "can come off as patronizing or standoffish." The same source notes あなた is the form used "to sound rational and professional when you communicate negative emotions, such as anger, hatred, hostility, or a sense of rivalry," so it carries distance.1 A parent who "usually calls you by your name" switching to あなた "makes the situation feel a lot more serious."1
The default is to leave the person out entirely. As one reference puts it, "We usually manage to communicate or tell a story without using a second-person pronoun in Japanese... We simply leave it out or call the person by their name, family role like 'mom,' or social role like 'doctor.'"1
Register: where omission tightens and where it loosens
Omission is heaviest in conversation. The balanced corpus that yielded the 37%-of-arguments-omitted figure spans text including the book domain, but conversational and casual registers drop the most.3 The register lever is visible in copula ellipsis: だ-drop is tied to casual "X is Y" statements without added modality, while formal or modality-laden statements restore the copula.4
Read against that lever, the direction is consistent. Casual speech elides the copula and arguments freely. Careful or formal speech tends to restore them for precision rather than the reverse. Writing that prizes unambiguous reference can reasonably be expected to re-specify more than conversation does, though the sourced contrast here is specifically the casual-versus-formal copula behavior.4
English contrast: why this feels alien
Low-context English forces the slots
English grammar requires an overt subject even when there is no referent to express. It inserts a dummy pronoun to fill the slot: "in the sentence 'It rained' the English pronoun 'it' is generally analyzed as a dummy pronoun, inserted to fill the subject position, but not referring to anything."7 The slot must be filled regardless of meaning.7
Japanese has no such requirement. A weather statement needs no dummy subject, and the predicate stands alone where English forces "it." Likewise, English requires an overt object pronoun where Japanese drops the recoverable object entirely.37
食べた。3
"(I/he) ate (it)." (English forces "I ate it"; Japanese needs neither slot.)
The learner consequence is a transfer error: English speakers over-produce 私は and それを because their first language drilled the obligatory slots. The corrective is the naturalness note. Spelling out a known subject "sounds unnatural" in Japanese and implies an unintended contrast.2
Good to know
Pitfall: over-using 私 and あなた
The single most common beginner tell is spelling out 私 ("I") and あなた ("you") when they should simply be dropped. Naming an obvious referent implies the hearer could not infer it. あなた in particular "can come off as patronizing or standoffish,"1 and overt known subjects "sound unnatural."2 Dropping them is the default, not the exception.
When asking a single, obvious listener what they want to eat, the wrong form spells out あなた (あなたは何を食べたいですか), which reads as cold or distancing. Drop it. 〜たい and the speech situation already recover the listener:
何を食べたいですか。2
"What would (you) like to eat?" (constructed example; the underlying naturalness claim is sourced)
Mnemonic: "if they already know it, leave it"
The recoverability principle reduces to a one-line test. If the element is "implied in the context" and "clear in context," drop it;2 if it "cannot be retrieved and correctly interpreted," keep it, because that omission would be ungrammatical.4 This phrasing is a learner's restatement of a sourced rule.
The term 省略 (shōryaku) and 言わぬが花
省略 (shōryaku) is the standard term for ellipsis and omission, and 主語省略 names the subject subcase.4 The cultural value attached to the unsaid is lexicalized in the proverb 言わぬが花 (iwanu ga hana). It literally means "not-saying is the flower" and is glossed "some things are better left unsaid, silence is golden."14 It captures the premium on the unstated that the high-context section explains.
Particles can drop too, but that is a different lane
This article covers dropping the argument itself: the noun and its case. A separate phenomenon drops the particle while the noun stays. That is the colloquial omission of は, が, or を in casual speech. The reference grammar treats subject omission and particle omission as interconnected in casual speech, noting that "omitting the subject and the particle は leads to the most natural sentence."2
They are distinct mechanisms: one removes a referent, while the other removes case-marking on a referent that remains. The particle-drop side belongs to the relevant particle material rather than here.
See also
- Dropped Subjects in Japanese: Pro-Drop Explained
- Scrambling and Word-Order Flexibility in Japanese
- Inversion (倒置): When Japanese Reverses Verb-Final Order
- Japanese Word Order: SOV and the Head-Final Principle