Japanese Filler Words and Hesitation Prosody: あの, えーと, まあ, and the Long-Vowel Stall
Japanese filler words and hesitation prosody are the conventionalised sounds and pitch shapes a speaker uses while planning the next stretch of speech: あのー, えーと, そのー, まあ, なんか, ちょっと, でー, やっぱり, and the stretched final vowel that turns any of them into a planner.12 A learner who masters the list but ignores the prosody can sound textbook-mechanical or register-mismatched in a job interview.
Overview
What counts as a filler (and what does not)
A filler, in Japanese linguistic terminology フィラー (firā), is "a semantically empty element of speech which fits a language-specific conventional phonetic form and delays… the transfer of the speaker's message."2 The broader term for the whole disfluency phenomenon is 言いよどみ (iiyodomi). It covers silent pauses, restarts, and repairs alongside the lexical fillers themselves.34
The older school-grammar label is 間投詞 (kantōshi). 大辞林 classes it as a subtype of 感動詞, the family of independent uninflected words that sit outside the predicate or modifier slots.5
あのー、ちょっとお願いがあるんですが。1
"Um, I have a small favour to ask."
Fillers are speaker-side floor-holding devices. Aizuchi, the listener-side backchannels like うん, へえ, and そうですか, are a different category produced by the person who is not holding the turn. Conflating the two is the most common error in popular treatments of Japanese conversation.13
A filler lets the speaker keep the floor while planning. A backchannel (aizuchi) lets the listener signal engagement. A hedge softens the strength of a claim. One form can do two of these jobs in different contexts (まあ is a hedge that doubles as a filler), but the three functions are conceptually distinct and the literature treats them separately.13
Why Japanese fillers exist: holding the turn while planning
Filled pauses are produced disproportionately before long or syntactically complex constituents in Japanese spontaneous speech, and native listeners use a preceding filler as a real-time cue that the upcoming phrase will be longer or more difficult to parse.6
The same predictive cue is weaker for L2 listeners: learners do not exploit Japanese filled pauses for upcoming-complexity prediction to the same degree as native speakers.6
Fillers also cluster after sentence-initial case particles and at major clause boundaries. The closer to the start of the sentence, the higher the rate of filled pauses immediately after a case particle, which fits a planning-load account of where speakers stall.7
In the Sadanobu and Takubo analysis, the two most-studied Japanese fillers split by cognitive operation. えーと signals that the speaker is "securing working space in mental memory" for search or calculation. あの(ー) signals "extraction of linguistic information from the knowledge base," meaning name retrieval or formulation of a phrase.1 The cross-linguistic baseline is similar: Clark and Fox Tree argue English uh and um are conventional words signalling minor versus major delay. The same "filler as conventionalised signal" view underwrites the Japanese literature.8
The eight-item core inventory at a glance
| Form | One-line gloss | Position in turn | Register tier |
|---|---|---|---|
| あのー (anō) | polite attention-getter, "about that…" | turn-opening; pre-FTA | workplace-safe19 |
| えーと / えっと (ētō / etto) | "let me think" | turn-opening or mid-turn | workplace-safe110 |
| そのー (sonō) | repair or refinement of an already-introduced item | mid-turn, post-mention | workplace-tolerant3 |
| まあ (mā) | hedge, downgrader, "well…" | turn-initial or mid | neutral3 |
| なんか (nanka) | vague-er, "like" | mid-turn, repeatable | casual; risky in formal settings3 |
| ちょっと (chotto) | softener, refusal marker, mild filler | pre-request or pre-refusal | neutral; context-dependent |
| でー / それでー (dē / soredē) | turn-continuation conjunction | turn-medial linker | casual to neutral3 |
| やっぱり / やっぱ (yappari / yappa) | "after all" recycled as thinking-out-loud filler | turn-medial | casual to neutral3 |
The four forms most often named as the core inventory in NINJAL and University of Tokyo corpus work are あの, え, ええと, and ま(あ). In Watanabe et al.'s elicited-preference study of presentation contexts, eight of ten native speakers chose ええと as their preferred thinking-time filler. あの and え were the next-ranked alternatives.9 The 国立国語研究所 Corpus of Spontaneous Japanese (CSJ) marks fillers and disfluencies with dedicated tags, and the inventory in the table is consistent with the CSJ filler tag set.1112
The filler inventory in detail
あの / あのー: the polite attention-getter
あの(ー) signals that the speaker is retrieving or formulating the linguistic content of an upcoming utterance, and it is the conventional Japanese opener before a request, an apology, or any other face-threatening turn.1
あのー、すみません、駅はどちらですか。1
"Um, excuse me, which way is the station?"
あのー、ちょっとご相談したいことがありまして。1
"Um, there's something I'd like to consult you about, if I may."
A common heuristic is that the longer the vowel, the more delicate the upcoming content: あの → あのー → あのーーー tracks rising face-threat. Filled pauses are documented as frequent in dispreferred responses and embarrassing remarks.1 That fits the heuristic, although a corpus-quantified curve mapping vowel duration directly onto face-threat weight does not appear to be in the published literature.13
The form has the same sounds as the addressee-distant demonstrative あの ("that [over there]"). What pulls the two apart in real speech is prosody, not segments: the demonstrative carries normal citation pitch and a short two-mora realisation, while the filler form is stretched and pitch-compressed.14 The same demonstrative-versus-filler split has been documented quantitatively for Mandarin 那 (nà) in the NINJAL–Academia Sinica collaboration on filled-pause typology.14
あのー、それなんですけど、来週でもいいですか。3
"Um, about that, would next week also be OK?"
あの(ー) sits in the workplace-safe tier alongside ええと across the polite-filler literature. It fits customer-facing and meeting speech without sounding casual. In Watanabe et al.'s preference data, あの ranks among the top three native-preferred fillers in presentation contexts.9
えーと / えっと: pure thinking time
えーと marks computational setup. The speaker is searching memory, performing a calculation, or constructing a response, and it is the closest functional equivalent of English um.18
えーと、確か三人来る予定です。10
"Um, three people are scheduled to come, if I remember right."
Turn-opening えーと carries an interactional load beyond mere planning. In conversation-analytic work it conveys "I find this question challenging, but I will make a maximum effort to respond," and establishes intersubjective alignment when an interaction has stalled.10
The short, clipped えっと and the long-vowel えーと or えーーっと sit on the same conventionalised continuum. Longer realisations are associated with heavier processing load. The general pattern, filled-pause duration scaling with the complexity of the upcoming phrase, is documented in Japanese spontaneous speech.67
えっと、それは答えにくい質問ですね。10
"Well… that's a difficult question to answer."
えーーっと、ちょっと待ってください。7
"Hmm, give me a moment, please."
えーと is the only filler in the inventory that the literature consistently flags as safe across the full register range, from job interview to peer-casual speech.19 The long-vowel variant えーーっと raises the perceived planning load but does not become rude as it lengthens. It only becomes more visibly hesitant.7
そのー: the second-mention or self-correction filler
そのー is the anaphoric filler, meaning it points back to something already mentioned. It leans toward repair or refinement rather than a fresh attention-getter. Where あの opens a new turn, そのー reaches back into the current one to pick up and reshape what was just said.3
In Ikeda's typology, そのー patterns with the speaker's effort to continue and refine an ongoing formulation, making it a planning marker rather than a hedge.3
田中さん、そのー、田中部長は今日はお休みです。3
"Mr Tanaka… I mean, Section Chief Tanaka is off today."
値段が、そのー、ちょっと高いかもしれません。3
"The price is, well, perhaps a little high."
Like あの(ー), そのー has a segmental twin in the demonstrative その ("that [near you]"), and the same pitch-flattening plus vowel-stretching distinguishes the filler use from the demonstrative.14
まあ: the hedging "well"
まあ softens, downgrades, or concedes. It can sit turn-initially ("well, I think…") or turn-medially ("it's, well, fine"). Ikeda groups まあ with the discourse-managing fillers that adjust the strength of the upcoming claim rather than buy planning time.3
大辞林 lists まあ both as an interjection of mild surprise and as a discourse softener equivalent to "well, sort of." The filler usage descends from the latter sense.5
まあ、それでもいいんじゃないですか。3
"Well, that's probably fine too, isn't it?"
まあ、結果としては成功でした。3
"Well, in the end it was a success."
A stack of まあ in a business setting comes across as non-commitment rather than polite softening. The hedge accumulates instead of cancelling out. Workplace registers tolerate one まあ in an answer; a wall of them sounds like the speaker is dodging the question.3
なんか: the casual "like"
なんか is a vagueness marker and hedge, and is the closest functional analogue to colloquial English like. It marks the upcoming item as approximate or as one example among many.3
There are two grammatical なんか. The noun-modifier なんか ("books and the like") is a quantifier-flavoured suffix that attaches to a noun.
The filler なんか is bleached of that quantifier content and inserted mid-clause as a planning or hedging device. The two share a form but not a function.3
なんか、最近よく寝られないんですよね。3
"I dunno, lately I can't really sleep well."
昨日、なんか変な夢を見ました。3
"Yesterday I had this kind of weird dream."
なんか、これ、なんか違くないですか。3
"Like, this, is this somehow off?"
Bunched repetition of なんか, two or three times in a short utterance, marks casual peer-aligned speech in Ikeda's typology and in popular workplace-Japanese guidance.3 The same pattern is widely cited in industry hiring blogs as a register clash in formal evaluations. However, corpus-grade quantification of that interview reaction is not in the published literature.
ちょっと: filler vs. softener vs. refusal marker
ちょっと sits at a three-way junction: it can be the literal adverb a little, a request softener, or a conventionalised polite-refusal marker. The three uses share the idea of "diminish the imposition." In practice, prosody and turn-position disambiguate them.
The refusal use is fully conventionalised: ちょっと trails off without ever stating the refusal in words. The speaker signals that an invitation is difficult to accept by saying only the softener.5
ちょっと聞いてもいいですか。5
"May I just ask something?"
今週はちょっと…。5
"This week is a bit… [difficult]."
あのー、ちょっと、お時間よろしいでしょうか。5
"Um, just briefly, would you have a moment?"
大辞林 separates these into related entries: quantity ("a small amount"), softener ("just a moment"), and "expression of mild reluctance, used to decline."5 A corpus-grade quantification of the three uses across registers is only partial in the published literature.
でー and conjunction-fillers
Elongated でー, それでー, and あとー are turn-continuation devices. They keep the floor while the speaker assembles the next clause, and they sit in the casual register tier.3
でー is the colloquial reduction of それで ("and so"). Both function as casual narrative connectors for stringing utterances together ("and then…").3
でー、その後どうなったんですか。3
"And, what happened after that?"
それでー、駅で待ち合わせしたんだけど。3
"And so, we met up at the station, but…"
あとー、もう一つお願いがあるんですけど。3
"Also, there's one more thing I wanted to ask."
These conjunction-fillers stack at clause boundaries the same way えーと stacks at planning points. The prosodic cue is the same lengthened final vowel that converts content words into fillers across the inventory.15
やっぱり as a filler
The content-word やっぱり means "as I thought" or "after all," and at JLPT N3 it appears as a confirmation adverb. In casual speech, it has been recycled as a thinking-out-loud filler. It is used at the moment of arrival at a conclusion rather than as a planning stall.3
大辞林 lists やはり / やっぱり as an adverb meaning "as expected, after all, all the same." The discourse-marker use is a pragmatic extension of the confirmation sense.5
やっぱり、コーヒーにします。5
"On second thought, I'll have coffee."
やっぱ、その案でいいと思うよ。3
"Yeah, I really do think that plan's fine."
やっぱり、最初の答えが正しかったみたいです。5
"As I thought, the first answer seems to have been correct."
The register is casual to neutral. やっぱり appears in spoken Japanese across most non-formal contexts but is rare in highly formal written or scripted speech.3
Hesitation prosody: the long-vowel stall
Mora-length stretching: the filler-making device
The signature acoustic property of a Japanese filler is lengthening of the final vowel beyond its canonical mora count: あの → あのー → あのーー. This is the device that converts the segmental form of a demonstrative or a conjunction into a planning marker.1614
The Japanese pillar treats the mora as a foundational unit in Mora vs. Syllable: Why Japanese Is Mora-Timed. Each added length unit in a filler is one mora's worth of duration. That is why writers represent the stretch with the chōonpu ー in katakana (アノー) and with iterative vowel kana in hiragana (あのお, あのー).17 The hiragana-versus-katakana split of that mark is covered in Long Vowels in Katakana: How the Chōonpu ー Works and Why Hiragana Doesn't Use It.
In CSJ data, filled-pause durations vary systematically with the complexity of the following constituent, with longer filled pauses preceding longer or more complex phrases.67
The hesitation pitch contour
Maekawa's F0 modelling of Japanese filled pauses finds that their fundamental frequency (pitch height) is predictable from the F0 of the preceding phrase and tends toward a low, relatively flat realisation. The proposed account is "tonal compatibility": filled pauses inherit a flattened version of the surrounding pitch span.16 This F0 work is framed in the NINJAL programme as an established line of investigation rather than a single-paper result.
The contour is distinct from the sentence-final fall used in declaratives and from the question rise. It is a sustained low-flat or gently falling plateau on the stretched vowel. Both the declarative fall and the question rise are treated in Japanese Sentence Intonation: Falls, Rises, ね, よ, よね. The filler plateau is the third shape that sits beside them.
Sherr-Ziarko's prosodic study of formality in conversational Japanese finds that formal speech is overall slower, quieter, and contains more pauses. Filler-laden hesitation patterns sit inside this broader prosodic-mitigation envelope rather than violating it.15
Position in the turn: opening, mid-turn, repair
Fillers cluster in three slots, each with its own prosodic signature.
- Turn-opening fillers (typically あの, えーと, まあ) carry a louder onset and signal an incoming turn.110
- Mid-turn hesitation fillers (えーと, なんか, でー) sit at lower amplitude and a more compressed pitch range, signalling in-progress planning.67
- Repair fillers (そのー, あー with cut-off and restart) appear at the point of self-correction.3
The Watanabe et al. case-particle finding fits the mid-turn pattern: filled pauses cluster after early case particles, where syntactic planning load is highest.7
The Sadanobu and Takubo mental-operation split maps onto turn position as well. あの(ー) fits best at the planning-and-formulation boundary, which is typically turn opening or pre-noun-phrase. えーと fits best at the search-and-calculation boundary, which is typically mid-turn before a content slot.1
How filler prosody differs from the demonstrative あの
The filler あのー and the demonstrative あの have the same sounds but different prosody. The demonstrative carries the citation pitch of a normal pre-noun adjectival. In standard Tokyo Japanese, that is the atamadaka shape, and it stays at a short two-mora realisation. The filler is stretched to three or more morae and pitch-flattened.1617
The full atamadaka contour is covered in Atamadaka (頭高): The Head-High Japanese Pitch-Accent Pattern.
NHK 日本語発音アクセント新辞典 records standard-Tokyo accent patterns for content-word あの as a pre-noun adjectival, and these citation patterns are precisely what the filler use flattens away.17
Same letters, different prosody, different meaning. If あの comes out short with a normal pitch jump, it is the demonstrative ("that thing"). If あの comes out stretched and pitch-flat, it is the filler ("um, about that…"). The lexical entry is one. The prosodic contrast does the disambiguating work.161417
The cross-linguistic FPRC project documents the same demonstrative-versus-filler disambiguation by F0 contour for Mandarin 那 (nà). The Japanese parallel is established as the pattern across the NINJAL programme rather than as a single published Japanese-specific quantification.14
Register and audience
The register tiers, mapped
| Filler | Job interview | Business meeting | Customer-facing | Neutral / friendly | Peer-casual |
|---|---|---|---|---|---|
| えーと / えっと | OK110 | OK9 | OK | OK | OK |
| あのー | OK1 | OK1 | OK | OK | OK |
| そのー | tolerated3 | tolerated3 | tolerated3 | OK3 | OK3 |
| まあ | caution3 | caution3 | caution | OK | OK |
| ちょっと (as filler) | tolerated5 | tolerated5 | OK | OK | OK |
| やっぱり | avoid3 | caution3 | caution3 | OK | OK |
| なんか | avoid3 | avoid3 | avoid3 | caution | OK |
| でー / それでー | avoid3 | avoid3 | avoid3 | caution | OK |
Workplace tolerance: the polite fillers that stay
The workplace-safe tier is small and stable: えーと, あのー, and そうですね. The last is conventionally classed as an aizuchi (listener backchannel) rather than a filler, but it is widely cited as the safest opener in an interview answer.19
These three retain a flat, low-amplitude prosodic profile that is consistent with Sherr-Ziarko's formal-speech pattern of slower, quieter, more hesitant production. They do not violate the prosodic mitigation that already characterises polite speech.15
Watanabe et al.'s preference data on academic presentations places えーと, え, あの, and ま as the top four fillers actually produced by Japanese presenters in semi-formal speech.9
Casual abundance: the fillers that pile up among friends
The casual tier, なんか, でー, やっぱ(り), あとー, is characterised by high density, repeated stacking, and conjunction-filler chains.3
The same items read as natural peer-casual speech when they appear at moderate density. When stacked densely, they become register markers for young or informal speech.3
An L2 speaker who borrows the なんか-stacked register before acquiring the prosodic and lexical co-occurrence constraints that license it tends to sound like a learner doing an impression, not a friend speaking naturally. The casual tier is not a shortcut; it is the same speech community's other register, with its own rules.63
Speech-style matching
The prosodic-formality literature treats register as a multi-dimensional bundle: F0 range (pitch range), speech rate, intensity, and pause frequency. Filler choice is one observable cue inside that bundle rather than an independent dial.15
Mismatching upward, by using only えーと and あのー with a casual peer, can read as cold or distant. Mismatching downward, by using なんか with a section chief, can read as presumptuous or unprofessional.3
The L2 failure modes
Sounding evasive: too many まあ and なんか
Heavy hedge-filler density, especially stacked まあ and なんか, reads as non-commitment in Japanese workplace registers. The hedges accumulate rather than soften.3
This is a register failure rather than a grammatical one. Each filler is well-formed; the density is what marks the speech as evasive.3
Sounding childish or unprofessional: なんか on repeat
Repeated なんか is the most-cited casual-register filler in workplace Japanese guidance. It patterns as casual peer-talk and is widely treated in industry blogs as immature or unprepared in formal evaluation. However, a corpus-grade study of that interview reaction is not in the published literature.3
The interview-board reaction is a register clash; the word itself is unobjectionable in peer speech. The fix is not to drop なんか from the active vocabulary but to keep it out of evaluative settings.3
Sounding mechanical: zero fillers
Classroom-perfect, zero-hesitation speech reads as scripted because native spontaneous speech contains filled pauses at predictable rates above zero. CSJ data shows native filled-pause rates of roughly 1.3 filled pauses per 100 words across the spontaneous-monologue corpus. Per-speaker rates range from approximately 0.2 to 6.6 per 100 words.9
The "1 to 3 fillers per minute" figure that circulates in pedagogy is a heuristic, not a measurement. Native rates vary substantially by genre, by speaker, and by speech style.9 Treat the figure as permission to use any fillers at all, rather than as a target.
Sounding like an L1 English speaker: mapping "um" 1-to-1 onto えーと
The Sadanobu and Takubo split (えーと = search or calculation; あの(ー) = retrieval or formulation) is the most-cited diagnostic for choosing the right filler for the right cognitive moment. A pre-name slot or pre-face-threatening opening calls for あのー. A pre-number, pre-date, or pre-recall slot calls for えーと.1
The 1-to-1 substitution fails because the Japanese inventory is functionally partitioned. えーと is for computational setup. あの is for linguistic retrieval and turn-opening. まあ is for hedging. なんか is for marking vagueness. A learner who reaches for えーと on every pause regardless of position or function sounds L2-flat, even when each individual えーと is well-formed.18
Good to know
Why あの, その, この, どの share a skeleton
The four pre-noun demonstratives in modern Japanese sit on the ko-so-a-do paradigm: speaker-near (この), addressee-near (その), distant (あの), interrogative (どの). The filler あのー extends the distant demonstrative ("that [over there]") into a discourse-management use ("about that thing I am trying to formulate"). This is the same kind of shift that took English well from an adverb into a hedge.54
The etymology is also what makes pitch and length the disambiguators: the demonstrative keeps citation pitch and short mora count, while the filler stretches and flattens.1617
Reaching for なんか in a job interview
A learner who opens an interview answer with なんか, 御社で働きたいと思いまして… intends tentative humility but lands on something that reads as casual peer-talk. The fix is to swap to a workplace-safe filler:
えーと、御社で働きたいと考えております。3
"Um, I am thinking that I would like to work at your company."
なんか is widely flagged in industry hiring guidance as a register clash for formal evaluations, and えーと is the workplace-safe equivalent for the same thinking-time slot.3
Stacking まあ in a business presentation
Piled-up まあ accumulates rather than softens, and a wall of まあ in a presentation reads as non-commitment or evasion. The workplace tier uses えーと and あのー freely and saves まあ for at most one occurrence per answer.3
"あの = retrieving a name; えーと = doing arithmetic"
The fastest disambiguation rule maps onto the Sadanobu and Takubo cognitive-operation split. A pre-name or pre-face-threatening-act slot takes あのー (linguistic retrieval). A pre-number, pre-date, or pre-recall slot takes えーと (computational setup).1 The mnemonic does not cover every edge case, but it gets the two most-used fillers right far more often than a coin flip.
Treating fillers as if they were aizuchi
A learner who produces うん or へえ while they are holding the turn has confused the two categories. While holding the turn, the correct devices are えーと, あのー, そのー. While the other person is holding the turn, the listener responds with うん, へえ, そうですか.
Fillers are speaker-side floor-holding devices; aizuchi are listener-side backchannels. The conflation is the single most common error in pop accounts of Japanese conversation.318
フィラー, 言いよどみ, 間投詞: three labels, three eras
The three Japanese-language labels are not interchangeable. フィラー (firā) is the linguistics-textbook loan, used in NINJAL and CSJ work, and applies specifically to conventional lexical fillers.1112 言いよどみ (iiyodomi) is the broader Japanese term for the disfluency phenomenon. It covers silent pauses, restarts, and repairs alongside fillers.34
間投詞 (kantōshi) is the older school-grammar label that lumps fillers with interjections (感動詞).5 Reference works will use whichever label fits their era; in modern linguistics, フィラー is the working term.
The "1 to 3 fillers per minute" pedagogy heuristic
The number is a heuristic, not a measurement. Native rates in the CSJ vary by genre (interview, monologue, casual chat) and by individual speaker, and per-speaker filled-pause rates in the corpus span roughly a thirty-fold range.9 Treat the figure as a permission to use any fillers at all, not as a target to hit.
See also
- Japanese Focus Prosody: Pitch Widening, Contrastive は, and Information Structure
- Japanese Questions Without か: The Rising-Intonation Question and the の Alternative
- Heiban (平板): The Flat Japanese Pitch-Accent Pattern
- Japanese Pitch Accent: A Complete Beginner's Guide
- Code-Switching (Style-Shifting): When Japanese Speakers Mix Registers
- How to Agree and Disagree Politely in Japanese: Hedging and Soft Disagreement