Grammatically unstable placeholders and morpho-syntactic remedies: evidence from East Asian languages

: When a communicator faces a word-formulation problem, they may use a placeholder (PH) such as whatchamacallit to avoid producing a target expression or to delay it. A PH is a dummy element used to ﬁ ll in the syntactic slot of a target item that a communicator is unable or unwilling to produce (e.g. due to memory lapse). Previous studies have generally been concerned with grammatically stable PHs (e.g. whatchamacallit , you-know-what ), ‘ grammatically stable ’ in the sense that they are acceptably used (as long as the syntactic, semantic, and pragmatic conditions are met) on their own, without a morphosyntactic aid. In this article, we describe ‘ grammatically unstable ’ wh -derived PHs in three East Asian languages: Japanese, Korean, and Mandarin. To give a speci ﬁ c example, typically, the PH use of Japanese dare ‘ who ’ is not fully acceptable unless it is doubled (i.e. dare ∼ dare ) or combines with a non- wh element (e.g. dare - sore , where sore is the medial demonstrative ‘ that ’ ). We show that the types of such remedial morphosyntactic operations vary from language to language and also within a language.


Introduction
In verbal interaction, a speaker may face a word-formulation problem when, for instance, they cannot remember an expression to refer to their intended entity. In such a situation, various types of phonological or lexico-grammatical strategies are available, including sound stretch, cut-off, interjective hesitators such as uh and um, and placeholders (PHs) (Fox 2010: 1). Of these strategies, our central concern is PHs. As has been amply described (e.g. Amiridze et al. 2010; see Section 2 for further references), PHs are well attested crosslinguistically, both in verbal and written registers. Seraku (2022a) defines PHs as in (1). (1) A PH is a dummy element with which a speaker/writer fills in the morphosyntactic slot of a target expression when they are unable or unwilling to produce it.
(2) [Naomi and Ken are flatmates. She wants to wash her clothes and sees a detergent on the shelf behind him. Naomi says to Ken.] Can you pass me that whatchamacallit?
[The speaker knows that the hearer's mother is very old, and asks whether she is fine.] mè 1 caw 4 ñang 2 bò 0 qan 0 -nan 4 juu 1 tii 4 mother 2SG still not thing-that PCL PCL 'Your mother's not yet PH (= senile), right?' (Enfield 2003: 108) [Lao] 1 In (2), the speaker is unable to recall the noun detergent and deploys the PH whatchamacallit instead. In (3), the speaker is unwilling to utter an expression denoting 'senile' because the use of such an expression may be impolite to the addressee. Such politeness consideration leads her to use the PH qan-nan (derived from qan 'thing' and nan 'that'). In (2)-(3), a PH substitutes for a target expression. As pointed out in Seraku (2022b), however, a speaker/writer does not always have in mind a specific expression or concept and uses a PH to refer to an arbitrary entity. In (4), takovat (< takovam, which derived from takova 'this kind of' and the verbaliser -m) refers to an arbitrary action. (4) [A Bulgarian professor working at a university in Japan talks with his colleague.] V Japonija prepodavatelite sa mnogo zaeti, zaštoto takovat v in Japan teacher.PL COP much busy.PL because PH in kampusa, takovat na konferencii, takovat vkǎšti i taka natatǎk. campus PH at conference.PL PH at.home and so onwards 'In Japan, professors are very busy because they PH on campus, PH at conferences, PH at home, and so on.' (Seraku 2022b: 446) [Bulgarian] There are three occurrences of takovat. Here, let us focus on the first occurrence. The syntactic slot before v kampusa 'on campus' requires an action-denoting verb, but the speaker is not in the position to produce any concrete verb because she is currently making a general statement. She avoids producing a specific verb with takovat, and by doing so, refers to an arbitrary action that a professor usually does on campus. This arbitrary-referential use of takovat is akin to the use of English capital letters in a contract form, such as in A and B agree on the terms and conditions stated below, where the capital letters substitute for a proper name that potentially occupies the syntactic slot and refer to an arbitrary person, company, etc. (Seraku 2022b: 434).
Following Seraku (2022b), we hold that (4) exhibits an arbitrary-referential use of the PH takovat rather than an expression distinct from the PH takovat. That is, we construe PHs broadly, encompassing not only instances such as (2)-(3) but also (4). For this purpose, we complement definition (1) with the finding of Seraku (2022b):

(5)
A PH is a dummy element with which a speaker/writer fills in the morphosyntactic slot of a target expression when they are unable or unwilling to produce it or when they intend to refer to an arbitrary entity.
This extended definition differentiates three types of PHs. For the sake of presentation, we call them a PH A , a PH P , and a PH ARB (with their plural forms being PHs A , PHs P , and PHs ARB ): (6) Three types of PHs a. The PH A refers to a PH that displays ability-related functions: a communicator is unable to produce a target expression, as in (2). b. The PH P refers to a PH that displays preference-related functions: a communicator is unwilling to produce a target expression, as in (3). c. The PH ARB refers to a PH that displays arbitrary reference functions: a communicator intends to refer to an arbitrary entity, as in (4).
In Section 2, we will compare the extended definition in (5) with several extant definitions of PHs and explicate why (5) particularly suits the present study.
In this article, we explore a variety of PH forms that derived from a wh-word in three East Asian languages: Japanese, Korean, and Mandarin Chinese. As will be argued in Sections 3-5, the wh-derived PHs in these languages are unique in that unlike whatchamacallit (2), Lao qan-nan (3), and Bulgarian takovat (4), they are grammatically unstable. As an illustration, consider (7), where the doubled (or reduplicated) form of dare 'who', i.e. dare∼dare, serves as a PH A . (7) [This is part of a travel essay. The author is drinking with his friend, Hattori, in a restaurant. A middle-aged woman who works there talks to Hattori.] Totsuzen, yo-nin-no obasan-no hito-ri-ga Hattori-kun-o suddenly four-CL-GEN middle.aged.woman-GEN one-CL-NOM Hattori-Mr-ACC yubisashi, "anta enka-kashu-no dare∼dare-ni niteiru-wane"-to point you enka-singer-GEN PH∼PH-DAT resemble-FP-COMP matashitemo geragera warat-ta. Boku-wa na-o iw-are-temo again uproariously smile-PST 1SG-TOP name-ACC say-PASS-even.if sono kashu-no kao-mo wakara-nai-node pokantoshi-teiru-to […] that singer-GEN face-even know-NEG-because have.a.blank.look-IPFV-then 'Suddenly, one of the four middle-aged women pointed to Hattori and laughed out loud again, saying "You look like PH, an enka (= Japanese traditional ballad) singer".' I had a blank look because even when I heard the singer's name, I didn't even know the face of the singer.' (BCCWJ) [Japanese: PH A ] As is clear from sono kashu-no kao-mo wakaranai 'I don't even know the face of the singer', the author does not know the singer's name and resorts to dare∼dare. In (7), if the doubled dare∼dare is replaced with the non-doubled dare, the sentence is less acceptable. In other words, acceptability is significantly enhanced by means of doubling. We call this type of PHs 'grammatically unstable PHs'. They contrast with 'grammatically stable' cases such as whatchamacallit, which may be acceptably used in its own right (as long as the syntactic, semantic, and pragmatic conditions are satisfied), 2 without any morpho-syntactic aid such as doubling.
Grammatically unstable PHs have been underdescribed in the literature. Here, we will reveal various ways in which the PH use of a wh-word is marginal (or unacceptable) but with its acceptability being improved by a morpho-syntactic operation. Before presenting the data, Section 2 reviews relevant previous studies. Sections 3-5 describe a wide range of wh-derived PHs in Japanese, Korean, and Mandarin, while Section 6 discusses some differences in usage between these three languages. By way of conclusion, Section 7 discusses implications and possible ideas for further research.

Previous studies 2.1 Definitions of placeholders
Since previous studies have put forward different (but partially overlapping) definitions of PHs, it is important to clarify our definition (5) against them.
Second, there are denotational variations, which are largely reduced to the narrower and the broader view of PHs. According to the broader view, PHs are construed in line with Seraku's (2022a) definition (1), more specifically in terms of a communicator's abilities and preferences. For example, Channell (1994: 162) states that PHs are used when a "speaker does not know/has forgotten name/noun" or a "speaker does not wish to use name/noun". Similarly, Jucker et al. (2003Jucker et al. ( : 1750 summarise the motives for using PHs as follows: "[o]n the one hand, the speaker might not be able to refer to the intended item by name […] On the other hand, the speaker might know the name but prefer not to use it". Cheung (2015: 275) lists typical contexts where a PH is utilised properly and divide them into two types: those where "the speaker uses the wh-placeholder to substitute a word or an expression that he or she cannot immediately utter" and those where "the speaker knows the target word or expression but chooses to withhold it".
According to the narrower view, the term 'placeholder' is reserved for what we call PHs A (Amiridze et al. 2010;Hayashi and Yoon 2006). Hayashi and Yoon (2006: 489) survey demonstratives in several languages including Japanese, Korean, and Mandarin and point out their "placeholder use", where a demonstrative "is used to hold the place for a momentarily unavailable word". At the same time, they point out that some of such demonstratives display the "avoidance use" too, where speakers "use demonstratives as substitutes for more specific lexical items […] because they feel inhibited from saying them openly since an explicit mention of them can be impolite, face-threatening, offensive, etc." (ibid.: 501). The avoidance use here corresponds to PHs P in our classification.
All in all, definition (1) underlies what extant studies regard as the basic functions of PHs (or the forms used as PHs, if the narrower view is adopted). Recall that definition (5) is broader than definition (1) in including cases where a PH refers to an arbitrary entity. Although the issue of arbitrary reference has not been addressed, except in Seraku (2022b), we include this point in our definition because, as will be shown in Sections 3-5, (i) PHs A , PHs P , and PHs ARB are all found in Japanese, Korean, and Mandarin and (ii) various morpho-syntactic operations remedy the acceptability of grammatically unstable PHs A , PHs P , and PHs ARB in these languages.

Functions of placeholders
In (6), PHs are classified into PHs A , PHs P , and PHs ARB . In some cases, a PH has a single type of function. Thus, whatchamacallit is a PH A , only related to the abilityrelated functions, and you-know-what is a PH P , only associated with the preference-related functions (Enfield 2003: 107). In other cases, a PH is related to multiple types of functions. Estonian see (which is traced to a pronoun, a demonstrative, etc.) acts as a PH A as well as a PH P (Keevallik 2010). As we will see, the wh-derived PHs in Japanese, Korean, and Mandarin may serve as a PH A , a PH P , as well as a PH ARB .
When a PH is employed for ability-related or preference-related reasons, there are various usage scenarios. A typical scenario for PHs A is (2), where the speaker fails to produce detergent due to memory lapse. In other scenarios, a speaker reading a passage aloud may encounter some illegible characters (Cheung 2015: 276). A typical scenario for PHs P is (3), where the speaker hesitates to utter a target form for reasons of politeness. In other situations, a speaker does not want to reveal their poor pronunciation (Channell 1994: 162) or feels reluctant to produce a socially sensitive term (e.g. a sexual term; see [8]). As will be illustrated shortly, the wh-derived PHs in Japanese, Korean, and Mandarin are also employed in a wide array of ability-related and preference-related situations.
Despite their usage-differences, what PHs A , PHs P , and PHs ARB all have in common is that they fill in a certain morpho-syntactic slot. Thus, whatchamacallit in (2) occupies the object NP slot, Lao qan-nan in (3) the nominal-predicate slot, and Bulgarian takovat in (4) the verbal slot. A PH may also occupy an adjectival, an adverbial, or even a clausal slot. A noteworthy case is kua in Ilocano (Austronesian), which may be used in any of these three types of slots (Rubino 1996: 657-659). This gap-filling property is essential because it fundamentally differentiates PHs from interjective fillers such as uh and um, the latter of which do not occupy a substantive morpho-syntactic slot. In our examples from Japanese, Korean, and Mandarin, wh-derived PHs only occupy a nominal slot. 3 (8) [On a Q&A site, a woman has written about her boyfriend, who is often naked when he is at home. She asked how she could get him to put on clothes, and someone replies.] Fuku-o ki-nai-nara kare-no ie-ni-wa ika-nai-to clothes-ACC wear-NEG-COND 3SG-GEN house-ALL-TOP go-NEG-COMP danko shuchoosu-bekidesu. The writer used nani presumably because he hesitated to produce a sexual term 'intercourse'. The PH ARB nani is illustrated in (9).

Forms of placeholders
(9) [The writer conveys that writing up a diary every day reduces his stress.] Boku jitsuwa mainichi nikki kai-teru-ndesu. 1SG in.fact every.day diary write-IPFV-MM.HON Kyoo-wa nani-ga at-ta-toka nanishi-ta-toka nani tabe-ta-toka. today-TOP PH-NOM happen-PST-etc. PH-PST-etc. PH eat-PST-etc. 'In fact, I write in a diary every day, like "A PH happened today", "I PH-ed", "I ate a PH", and so on.' (SNS post) [Japanese: PH ARB ] The first nani in (9) refers to an arbitrary event, whilst the last refers to an arbitrary food. Also, nanishi (< nanisuru), derived from the nominal PH nani and the light verb -suru 'do', refers to an arbitrary action. These arbitrary references are crucial for explaining what kinds of topics the writer writes about in his diary. 5 Since the PH use of nani 'what' is described in Seraku (2022b), we concentrate here on other wh-words: dare 'who', itsu 'when', and doko 'where'. In (10), doko refers to an arbitrary place. (This example also contains the PH ARB nani∼nani, for which see Seraku [2022b].) 5 An anonymous reviewer suggested that it would be beneficial to clarify differences between PHs ARB and indefinite pronouns. First, unlike indefinite pronouns, PHs ARB may be verbal, as in nanisuru in (9). Second, the motive for using a PH ARB is distinct from that for using an indefinite pronoun. In (9), the writer uses nani to fill in the subject slot of at-'happen' and the object slot of tabe-'eat' since he does not have in mind a concrete noun denoting an event or food. This filling-in process is something we do not see in the use of indefinite pronouns. Third, unlike indefinite pronouns, PHs ARB usually appear in sentences describing a state-of-affairs that is spatiotemporally uninstantiated. In (9), the writer does not refer to a spatio-temporally instantiated 'eating' event. (10) [On a Q&A site, a beginner at PC asked where document files, once created, should be saved on the PC. Someone replies to this question.] Jibun-no pasokon-nara dokonidemo tsukut-tekudasai. Tadashi […] self-GEN PC-COND anywhere create-please but nani∼nani-wa doko-to kime-teoi-tahoogayoi-desu-ne.

PH∼PH-TOP
PH-QUOT establish-in.advance-had.better-HON-FP 'If you use your own PC, you can create a folder anywhere (and save a file there). But you should establish your own rule like "I always put a PH (= file) in a PH (= folder)." (BCCWJ) [Japanese: The writer is giving the questioner a tip for managing files on a PC, and for the sake of illustrating this tip, refers to an arbitrary folder by deploying doko.
Whilst (10) is fully acceptable, the PH use of dare, itsu, or doko is not fully acceptable in many other cases. In what follows, we will illustrate such cases, revealing two types of morpho-syntactic operations to enhance the acceptability of a PH form: doubling and combinations with a non-wh element.

Doubling
Our BCCWJ search 6 detects 50 clear occurrences of the doubled form dare∼dare (cf. dare 'who') that may be seen as a PH. Since an example of the PH A dare∼dare has already been given in (7), we here present an example of the PH ARB dare∼dare in (11). (11) [On a Q&A site, someone asked about tips for home buying. Someone else replies.] Ginkooman-kara osusume-no gyoosha-o kiki sono gyoosha-e banker-from recommendation-GEN company-ACC hear that company-ALL If you ask a banker about their recommended company and go there saying "I came here since Mr./Ms. PH referred me to you.", the person in charge will do their best for you, I suppose.' (BCCWJ) [Japanese: PH ARB ] The speaker uses dare∼dare to refer to an arbitrary banker. In (11), replacement of dare∼dare with dare renders the sentence less acceptable though not completely unacceptable. That is, doubling stabilises the PH ARB use of dare.
The BCCWJ contains 21 PH cases of itsu∼itsu (cf. itsu 'when') and 49 PH cases of doko∼doko (cf. doko 'where'). For reasons of space, we only present an example of itsu∼itsu. (12) [In this interview article, the speaker, the governor of a prefecture, explains the situation of nuclear power generations in the 1980-1990s.] 93-nen-ni-wa, Tooden-kara shiyoozumi kaku-nennryoo-o hokansuru 93-year-in-TOP T.E.P.C.-from used nuclear-fuel-ACC store puuru-o genpatsu-no koonai-ni zoosetsus-ase-tekure-to pool-ACC nuclear.power-plant-GEN premises-in expand-PASS-BEN-COMP iw-are, kuni-ga itsu∼itsu mochidasu-to hoshoosuru-koto-o say-PASS nation-NOM PH∼PH carry.out-COMP promise-NMNZ-ACC jooken-ni mitome-mashi-ta condition-at approve-HON-PST 'In 1993, Tokyo Electric Power Company asked us to expand the pools for used nuclear fuels in the premises of the power plant. We approved it on condition that the Japanese government would carry them out in PH (= year).' (BCCWJ) [Japanese: PH A /PH P ] The speaker may not have remembered the exact date when spent nuclear fuels were promised to be extracted from the storage pools, or he may have thought that such exact information was not important for the current topic of conversation. That is, itsu∼itsu in (12) may be viewed as a PH A or a PH P , depending on the interpretation. If itsu∼itsu is replaced with itsu in (12), the sentence becomes unacceptable, which indicates that doubling stabilises the PH A /PH P use of itsu.
The doubled forms such as dare∼dare, doko∼doko, and itsu∼itsu are treated in Sudo (2008Sudo ( , 2013 and Kudo (2020). Sudo offers a formal-semantic analysis of the doubled forms and their quantificational properties, whereas Kudo considers the doubled forms in connection with language change. In these studies, however, doubling is not seen as a remedial operation to stabilise the PH use of a wh-word.

Combinations with a non-wh element
Doubling is a systematic operation in that it applies to dare, doko, and itsu across the board. By contrast, as will be pointed out below, combinations with a non-wh element are idiosyncratic: (i) they only apply to dare and doko, and (ii) there are only a few types of combined elements.
The first type of combined element is a medial series of demonstratives: sore 'that' and soko 'there'. 7 We found the composite forms daresore (cf. dare 'who') and dokosoko (cf. doko 'where'). The BCCWJ has 48 cases of daresore and 62 of dokosoko. In (13), daresore and dokosoko appear as PHs ARB .
thing-NOM yoku aru. often exist 'Even when one gossips about something, the following thing often happens: when one hears that "Mr. PH comes from PH," he says something like "Ah, that's why" and suddenly admires the person.' (BCCWJ) [Japanese: PH ARB ] The writer elaborates on her claim (i.e. 'Japanese people often admire those who were brought up well.') by providing a hypothetical exchange. In this exchange, the writer needs to talk about a person and their birthplace without specifying them; for this purpose, she uses daresore and dokosoko. In (13), replacement of daresore with dare lowers the acceptability of the sentence, whilst replacement of dokosoko with doko does not significantly alter acceptability.
These composite forms may also act as a PH A or a PH P . Consider (14): The writer read the diary of Ozu, but it seems they do not recall its details, including the name of the student who got suspended or expelled from school. Or, even if the writer recalls the student's name, they may have thought that such details are not relevant. (14) is less acceptable if daresore is replaced with dare.
The second type of combined item is toka, which is translated into 'etc.' or 'something like that'. The BCCWJ has only three clear instances of daretoka, one of which is (15). (15) [This is a scene from a novel. A male working at a swimming school had an affair with one of his students.] uwasa-ga tat-te-ta-nda-sooda. go-IPFV-CVB rumour-NOM circulate-IPFV-PST-MM-HSY 'I heard that the guy was having an affair with a married woman who regularly came to the swimming school. According to PH who lived in the same government facility as Shimao and regularly went to the school, the rumour was circulating in the school.' (BCCWJ) [Japanese: PH A /PH P ] The speaker may not know the name of the person who lived in the same government facility as Shimao, or perhaps cannot recall the person's name. Alternatively, the speaker recalls the name but thought that it did not count as essential information, compared with the more important information that the guy had an affair with his student. In (15), the combination with toka is indispensable because replacement of daretoka with dare results in unacceptability.
The BCCWJ contains three PH cases of dokotoka. Consider first (16): (16) [This is part of the Diet Record. In the excerpt, a politician points out a disadvantage of classifying inmates into groups (e.g. a group of sexual offenders, a group of alcoholics) in a prison and educating each group separately.] Mata dokotoka-keemusho-shusshin-no hito-wa further PH-prison-from-GEN person-TOP seehanzaisha-da-zo-toiuyoona sexual.criminal-COP-FP-like retteru-o har-are-ru-toiuyoona demeritto-mo aru-to omoi-masu. label-ACC attach-PASS-PFV-like demerit-also exist-COMP think-HON 'Further, I think there is also a demerit, such as the one that a person discharged from the PH-prison is labelled as a sexual criminal.' (BCCWJ) [Japanese: PH ARB ] The speaker is talking about a disadvantage of classifying inmates in terms of crime types, and illustrates this opinion with a hypothetical statement, where they refer to an arbitrary prison. (16) is unacceptable with the bare form doko.
In (17), dokotoka may be regarded as a PH A or a PH P .
(17) [This is a scene from a novel. The second son of a married couple came to see them and told them that his daughter would take part in a game of mini basketball.] Ashita, dokotoka-no shoogakkoo-de minibasuketto-no tomorrow PH-GEN elementary.school-LOC mini.basketball-GEN shiai-ga aru-toiu. game-NOM exist-HSY 'According to (our son), there will be a game of mini basketball in an elementary school of PH (= area) tomorrow.' (BCCWJ) [Japanese: PH A /PH P ] Note that the speaker is reporting what he heard from his son. It is likely that his son told him about the area where the school is situated, but he may have forgotten it, or even if he remembers it, he may have thought that it is not worth mentioning.
In (17), use of toka is essential; if doko is used on its own, the sentence is unacceptable.

Summary
This section has considered the PH uses of three wh-words: dare 'who', doko 'where', and itsu 'when'. Unlike nani 'what' in (8)-(9), the PH uses of these whwords are generally marginal or unacceptable. Whilst it is a residual issue to spell out what factors govern their acceptability patterns, 8 we have shown that when the PH use of a wh-word is not fully acceptable, a certain morpho-syntactic operation may improve the acceptability, as delineated in Table 1.

Wh-derived placeholders in Korean
As argued in Section 3, the PH use of Japanese wh-words other than nani 'what' is limited. In Korean, wh-words are more freely used as PHs. Here, we focus on nwukwu 'who', mwe 'what', eti 'where', and encey 'when'. 9 Consider (18)  pleasure-NOM come-NEG-HON 'Having a sexual relationship with someone you just met may be a one-night stand, and well…, you may be stimulated by the sensual PH (= sex), but you will never be mentally satisfied with it.' (Yonsei Corpus) [Korean: The speaker talks about intercourse and seems to have produced mwe to mask a sexual term. Note that mwe is marked with the topic particle -nun. In Korean, whwords cannot be topic-marked in general (except for rhetorical questions; see Jung [2017]). It is thus clear that mwe does not act as an interrogative wh-word here.
In (19), encey 'when' is exploited as a PH A or a PH P .
[This is a scene from a novel. People on a swaying ship are overwhelmed by the fear of capsizing.] Seykyey choytay-uy hohwa yekayksen-to chimmolha-nun tey-nun world biggest-GEN luxury cruise.ship-even sink-ATT NMLZ-TOP pyel-swu-ka eps-ess-ta.
[…] amwuli extraordinary-means-NOM don't.exist-PST-DECL no.matter.how ancenhan pay-laciman… ha-nun uykwusim-i meli-lul naymi-nun safe ship-but say-ATT doubt-NOM head-ACC protrude-ATT kes-i-ess-ta. siktang aph-ul cina-ca cwupang salam-i MNLZ-NOM-PST-DECL restaurant front-ACC pass.by-as kitchen person-NOM cen-ey encey-nun pay-ka 40-to-na kiwule-cy-ess-nun tey.to before-at PH-TOP ship-NOM 40-degree-even tilt-become-PST-ATT but kkuttek-i eps-ess-ta-ko salam-tul-ul ansimsikhy-ess-ta. It is probable that the person working in the galley mentioned a specific year (or perhaps month or date) when the ship listed, but the speaker may not have remembered the exact year or may have thought that this information is less important than the information that the ship's listing did not result in a disaster.
The Korean wh-words nwukwu 'who', mwe 'what', eti 'where', and encey 'when' may also be used as PHs ARB . Consider (20) for the PH ARB encey and the PH ARB nwukwu.

TOPIC-TOP limitless
Grammatically unstable placeholders 'Then, they talk about their daughters, brag about the clothes that their daughters-in-law bought for them, and talk about their jewellery. And, again, they talk about the PH (= date) of the party to celebrate the 70th birthday of PH (= person), and they say that they will see the granddaughter-in-law of PH (= person). The topics of conversations of the elders are limitless.' (Yonsei Corpus) [Korean: PH ARB ] The writer explains in general terms what kinds of topics elderly people usually talk about. Since the writer does not have in mind any specific elderly people, neither encey nor nwukwu is used to refer to a specific individual.
In the literature of Korean linguistics and Korean dictionaries, it has been wellknown that wh-words are used for non-interrogative purposes (see I [2017]). The wh-words in (21)-(22) may be interpreted as a PH P or a PH ARB depending on context. For example, in (22), eti is seen as a PH P if we assume that the speaker presumes that the hearer would not know the place name even if they uttered it; eti is seen as a PH ARB if we assume that they in fact do not need to go anywhere but says (22) to decline the hearer's invitation to lunch. In this way, the interpretation of a PH crucially requires contextual information, but such a context-sensitive analysis has not yet been provided in the dictionary or previous studies in the tradition of Korean linguistics. It is, therefore, fair to say that Korean wh-derived PHs have not been sufficiently described.
As argued in Section 3, Japanese features doubling and combinations with a non-wh item as remedial operations to stabilise the PH use of wh-words. Of these two types of operations, doubling is attested in Korean, too. Consider (23): [This is part of a magazine article. A prosecutor is expressing his opinion about what will happen if the police are granted independent investigatory power.] Han socang kemsa-nun "[…] kyengchalsecang-eykey cinachin a chief prosecutor-TOP police.chief-to excessive kwenhan-i cipcwung-toyn-ta. Kyengchal cocik-uy kwuco.sang power-NOM focus-become-DECL police organisation-GEN structurally secang-i 'nwukwu∼nwukwu-lul senchehay cwu-la'-ko cisiha-myen chief-NOM PH∼PH-ACC go.easy BEN-IMP-COMP order-COND haytang cikwen-i kepwuha-ci.moshal kes-i-ta.
[…]"-ko in.question staff-NOM refuse-cannot NMLZ-COP-DECL […]"-COM moksoli-lul nophy-ess-ta. voice-ACC raise-PST-DECL 'A chief prosecutor raised his voice, saying "[…] Power is overly concentrated in the chief of a police station. Given the structure of the police organisation, if the chief orders to his staff "Go easy with (= commute) the crime of PH (= person)," they will not be able to refuse.".' (Sejong Corpus) [Korean: PH ARB ] The point of this excerpt is that authority is overly concentrated in the chief of a police station, and this general point is illustrated with the hypothetical utterance including nwukwu∼nwukwu. Therefore, nwukwu∼nwukwu does not refer to any specific individual. When nwukwu∼nwukwu is replaced with nwukwu, (23) is less acceptable with the PH ARB reading. 11 Thus, doubling is an essential operation to stabilise the PH ARB use of nwukwu here. 12 11 An anonymous reviewer pointed out that the non-doubled form nwukwu is acceptable in (23). The second author of this article, a native speaker of Korean, also feels that the clause secang-i 'nwukwu-lul senchehay cwu-la'-ko cisiha-myen 'if the chief orders to his staff "Go easy with the crime of PH"' itself is acceptable, but nwukwu in this clause seems to be more naturally interpreted to refer to a specific person, rather than an arbitrary person. We thus hold that (23) is less acceptable (if not unacceptable) with the PH ARB reading when nwukwu is involved. We consulted three Korean speakers (linguists), and they largely agreed with these points. 12 In Korean, doubling of a wh-word may lead to a wh-question with a multiple-entity reading. In (i), the speaker presupposes that more than one person came to the meeting. In (i), nwukwu nwukwu is not treated as a PH because it triggers a wh-interpretation. In Japanese, this use of wh-words is possible only in some dialects (Kudo 2020: 30). In Mandarin, the multipleentity interpretation of a wh-question is not possible, according to the third author (born in Nanjing) of the present article.
[…] annyenghasipnikka cehuy information-record-NOM come.out-HON how.are.you.HON 1PL eti∼eti-nun chincelhan sepisu-lo kokayk yelepwun-kkey PH∼PH-TOP considerate service-INS customer everyone-to potap-khoca nolyekhako.isssupnita. respond-intend make.an.effort.HON 'When we call a credit card company to ask about PH (= thing), a person does not take up the call but a recorded message is played, together with music. "How are you? We PH (= company name) are making an effort to be considerate in our service to customers.".' (Sejong Corpus) [Korean: PH ARB ] (25) [This is part of a magazine article. The writer is disgusted with those Koreans who urge the necessity of internationalising Korea.] Oykwuk-eyse com sal-ten salam, calnan.chekhanun ilpwu foreign.country-in a.bit live-PST.ATT person pretentious part kwukmin-un "Hankwuk-un mwe∼mwe ttaymwuney an.toynta"-ko citizen-TOP "Korea-TOP PH∼PH because.of not.good"-COMP ha-nta. "Hankwukin-un ilen nappun cem-i iss-ta"-ko-to say-DECL "Koreans-TOP this bad point-NOM exist-DECL"-COMP-also ha-nta. say-DECL 'Those Koreans who lived abroad for a while or some pretentious people in Korea say "Korea is not good because of PH." They also say "The Koreans are wrong in such-and-such respects.".' (Yonsei Corpus) [Korean: PH ARB ] (24) is rendered unacceptable with the non-doubled form eti. As for (25), it becomes less acceptable with the non-doubled form mwe. Thus, as in the case of Japanese, doubling may serve as a remedial operation for wh-derived PHs ARB in Korean.

(26)
[This is a scene from a novel. They are talking about a certain Korean company.] Hayoy kensel-ul ha-nuntey, yenge-lo hoysa ilum-i foreign.country construction-ACC do-CONN English-INS company name-NOM "Kholia mwe∼mwe" kulay-ss-ketun. Kulenikka ceccok-eyse mwusun "Korea PH∼PH" like.that-PST-FP therefore there-at well kwukyeng kiepchey kathun kes-ulo ala-ss-na.pwa. state.run enterprise like thing-as think-PST-seem 'When (the company) ran a construction business abroad, its English company name was "Korea-PH." So, it seems it was seen as something like a state-run enterprise.' (Yonsei Corpus) [Korean: PH P ] It is highly likely that the writer knows the exact name of the company. But what matters in the present context is that the company's name includes Korea, and the remaining part of the name is unimportant. This seems to be why mwe∼mwe was used as a PH P . (26) becomes less acceptable if mwe∼mwe is replaced with mwe. 13 As far as this example is concerned, then, doubling stabilises the PH use of mwe.
te.isang changphihay-se pethi-ci.mosha-ko ttal-ul any.more.NPI be.embarrased-CVB bear-cannot-CVB daughter-ACC Gwangneung-ey iss-nun cakun imo-eykey ka-se Gwangneung-in exist-ATT little maternal.aunt-to go-CVB sal-key hayyahay-ss-ta. live-CAUS must-PST-DECL 'The scene where they did a (sexual) thing seemed to come to light to the children in the town, and at some point, doodles started to appear everywhere in their neighbourhood, saying PH (= person) did a PH (= thing) with PH (= person). And soon after, the rumour about them spread to every neighbourhood. Her parents were so embarrassed that they had no choice but to make her live with her aunt in Gwangneung.' (Sejong Corpus) [Korean: PH P ] There are two occurrences of nwukwu∼nwukwu (cf. nwukwu 'who'); one refers to a female (i.e. 'daughter'), and the other to her boyfriend. (27) remains acceptable even when the doubled form is replaced with its non-doubled version nwukwu. That is, nwukwu in (27) may act as a PH P on its own without the aid of doubling.

(28) [This is a scene from a novel. A man is speaking of his wife.]
Anay-nun wul-taka wus-ess-ta. wife-TOP cry-and.then laugh-PST-DECL Wul-taka wus-umyen eti∼eti-ey thel na-nta-ten.tey. In Korean, there is an old saying: 'If you cry and then laugh soon, hair grows on your buttocks.' This saying is mockingly uttered to those who have emotional ups and downs. As the buttocks are a socially sensitive part of the body, the writer avoids mentioning it. It makes no difference to the acceptability of (28) whether eti∼eti is replaced with eti or not. It is at present not clear what conditions affect the PH use of a wh-word (see footnote 13), but in some cases, a bare wh-word cannot act as an acceptable PH and requires doubling as a remedial operation.
In sum, the preceding discussion has argued for the following points: -Nwukwu 'who', mwe 'what', eti 'where', and encey 'when' may act as a PH A , a PH P , or a PH ARB on their own, more freely than the Japanese counterparts (except for nani 'what').
-In some cases, the PH uses of these Korean wh-words are marginal or illegitimate and need to be stabilised by the doubling operation. -Combinations with a non-wh item (e.g. demonstrative), which serve as a remedial operation for Japanese wh-derived PHs (except for nani 'what'), are not attested in Korean.

Wh-derived placeholders in Mandarin
As illustrated in Cheung (2015) In Mandarin, the PH use of a wh-word is generally more acceptable when it combines with the demonstrative na 'that' (and optionally the general classifier ge). 14 In (29), the PH sequence na ge shenme substitutes for a word meaning 'router', which temporally eludes the speaker. What has not been discussed in Cheung (2015) and other studies (Hayashi and Yoon 2006;Lee et al. 2017;Yu and Wu 2015) is that Mandarin wh-derived PHs may be doubled. 15 In this section, we show that the doubled forms are also grammatically unstable and need to combine with the demonstrative na (and optionally with the classifier ge). To this end, we carried out a questionnaire survey, asking 29 participants to judge whether shei∼shei (cf. 'who') and shen-me∼shenme (cf. 'what') are acceptable on their own or need to be supplemented by 14 The Mandarin demonstratives are dichotomous, involving the proximal zhe 'this' (and its plural form zhexie 'these') and the distal na 'that' (and its plural form naxie 'those') (Wang 1944: 34-59). For a recent account of Mandarin demonstratives, see Lin (2020: Ch. 4-5). 15 Cheung (2015: 305-306) briefly notes that Mandarin wh-derived forms may be doubled, but all his examples, such as (i) below (where nali derived from 'where'), are most naturally construed as cases of PHs ARB .

(i)
Meimei rang-zhe yao qu nali∼nali guangjie. sister demand-PROG want go PH∼PH shopping 'My sister keeps demanding that she wants to go shopping here and there.' (Cheung 2015: 305) [Mandarin: PH ARB ] According to Cheung, doubled forms in examples like (i) are acceptable on their own; in fact, their acceptability is lowered if they combine with na or ge. Since our primary interest lies in grammatically unstable forms and remedial operations, we do not analyse the doubled wh-derived PHs ARB such as (i) above.
na and/or ge. 16 The types of a PH sequence concerning us here, therefore, are as follows: na + ge + wh-wh: na ge shei∼shei, na ge shenme∼shenme na + wh-wh: na shei∼shei, na shenme∼shenme ge + wh-wh: ge shei∼shei, ge shenme∼shenme wh-wh: shei∼shei, shenme∼shenme It should be noted that the doubling operation here is not seen as a remedial operation because all of the acceptable examples to be presented below remain acceptable when the doubled PH form is replaced with its non-doubled counterpart.

Shei∼shei
We start with shei∼shei (cf. shei 'who'). The participants were presented with the three sets of data given in (30) These examples are cases of a PH A or a PH P , but not a PH ARB . First, we asked the participants to interpret them in two contexts suitable for PH A /PH P readings: (i) the speaker cannot recall a target form and (ii) the speaker, though they are aware of a target form, prefers not to verbalise it. Second, Cheung (2015: 305) states that a doubled wh-word is only licensed in embedded indirect speech contexts.
In his examples (see footnote 15), doubled wh-words indeed occur in such contexts, but his examples are (at least, most naturally) seen as cases of PHs ARB . In contrast, the wh-words in (30)-(32) appear in the matrix clause. If they were PHs ARB , (30)-(32) would be unacceptable, contrary to fact. This indicates that when a doubled wh-word behaves as a PH ARB , its presence is restricted to the embedded indirect speech contexts, but when it behaves as a PH A or a PH P , the restriction does not apply. The acceptability patterns of (30)-(32) are summarised in Tables 2-4, where the 'ability-related context' refers to a context where the speaker cannot recall a target form (PH A ), and the 'preference-related context' to one where the speaker prefers not to verbalise a target form (PH P ). The participants were asked to judge the acceptability of each sentence on a scale from 1 to 3; with 1 being unacceptable (= *), 3 acceptable (= ok), and 2 somewhere in between (= ?). In all of (30)-(32), use of shei∼shei is more acceptable when it combines with the demonstrative na. This echoes the pattern of non-doubled PH forms; see the paragraph following (29). 18   Also, note that the bare classifier ge appears in the subject position in (30c), (31c), and (32c). An anonymous reviewer pointed out that a bare classifier generally cannot appear in the subject position and that this may be why almost all participants judged (30c), (31c), and (32c) to be unacceptable. We wish to take the present point into consideration when we conduct a more largescale experiment in future work.

Shenme∼shenme
Let us turn to shenme∼shenme (cf. shenme 'what'). The 29 participants were presented with the three sets of data in (33) The results of the acceptability judgements are presented in Tables 5-7. Since shenme∼shenme refers to a non-human entity, we prepared example sentences in which shenme∼shenme functions as an object (or part of an object). The predicate in (35) denotes an event, and the predicates in (33)-(34) (one is verbal, and the other non-verbal) denote a state.

Grammatically unstable placeholders
Thus, in (33)-(35) too, use of shenme∼shenme is generally more acceptable when it co-occurs with the demonstrative na. Still, use of shenme∼shenme is not fully unacceptable without na in (33c)-(33d): more than half of the 29 participants judged (33c) to be acceptable. This may have to do with the use of ge to mark indefiniteness in an object position. As illustrated in (36), ge may realise an indefinite, non-specific reading when (i) it is used for the object NP of a transitive verb and (ii) the transitive verb precedes ge (Lu 1984); see also Lu (2017).

(36)
Wo jintian zhi chi le ge pingguo. I today only eat ASP CL apple 'I only ate an apple today.' [Mandarin] In (33c), the PH sequence ge shenme∼shenme itself substitutes for a specific proper name such as Chanel, but the whole object NP ge shenme∼shenme bao refers to an indefinite, non-specific entity: a bag of Chanel. We presume that the participants who judged (33c) to be acceptable are those who construed it with this indefinite, non-specific reading, whereas the participants who did not judge (33c) to be acceptable are those who did not obtain this reading. Note that this indefinite reading is grammatically ruled out in (34c), where a transitive verb is absent, and also in (35c), where the transitive verb na 'bring' is present but does not precede ge.

Summary
We have revealed that (i) contrary to what Cheung (2015) suggests, Mandarin has the doubled wh-derived PHs A /PHs P , (ii) in general, they are grammatically  unstable unless they combine with na 'that', and (iii) when they are legitimately used without na, certain grammatical factors seem to be at work (e.g. the use of ge to mark indefiniteness in an object position; see also footnote 18). It should be recalled that unlike Japanese and Korean, doubling is not a remedial operation in Mandarin; the non-doubled forms, shei and shenme, may stably act as PHs on their own as long as they combine with na. What is seen as a remedial operation in Mandarin, then, is the combination with na, both for the doubled and the nondoubled forms. Despite these findings, our analysis is limited in that we have only considered shei∼shei and shenme∼shenme and the results of the survey have not been analysed in statistical terms. In future work, we hope to conduct a largerscale survey with a solid statistical analysis.

Cross-language and intra-language considerations
We are now in the position to present cross-language and intra-language comparisons. From a crosslinguistic point of view, different languages make use of different remedial operations: -Japanese: Doubling; Combination with a medial-series demonstrative or toka 'etc.' -Korean: Doubling -Mandarin: Combination with the demonstrative na In Mandarin, doubled PH forms are attested, but doubling is not viewed as a remedial operation because non-doubled PH forms themselves are licit (as long as the demonstrative na is present). This contrasts with Korean, where non-doubled PH forms are legitimate in many but not in all examples (e.g. [24]-[26]). As far as PHs are concerned, then, the stabilising power of doubling differs from language to language in the following manner: -Japanese: Doubling (or a combination with a non-wh item) is generally required to stabilise the PH use of a wh-word such as dare 'who', itsu 'when', and doko 'where'. The exception is nani 'what,' which may be productively used as a PH on its own. -Korean: The PH use of a wh-word such as nwukwu 'who', mwe 'what', eti 'where', and encey 'when' is widely observed, but in some cases, doubling is required to stabilise the PH use of these wh-words. -Mandarin: There are doubled PH forms based on shei 'who' and shenme 'what', but their non-doubled PH forms are also acceptable (as long as the demonstrative na is present). That is, the PH use of these wh-words is acceptable without doubling.
Another notable point about the remedial operations is that a combinatory operation is only available in Japanese and Mandarin. In Mandarin, the type of combined item is systematic compared with Japanese. In Mandarin, shei∼shei and shenme∼shenme always combine with the same element, the demonstrative na 'that' (and optionally with the classifier ge), while in Japanese, there is a variety of elements that can be combined with a wh-word: sore 'that' for dare 'who', soko 'there' for doko 'where', and toka 'etc.' for dare 'who' and doko 'where'. Moreover, this combinatorial option is not available for all wh-words; as far as our consulted corpus is concerned, itsu 'when' does not combine with a non-wh element.
Finally, a bare wh-word may serve as a PH most robustly in Korean. As stated in Section 4, each of nwukwu 'who', mwe 'what', eti 'where', and encey 'when' may stand as a PH in its own right in many cases. In contrast, the wh-words in Japanese and Mandarin generally require a remedial operation, except for Japanese nani 'what', which is used productively without any morpho-syntactic support. Crosslinguistically, then, we may conclude that the grammatical stability of the PH use of a bare wh-word decreases as follows: Korean > Japanese (>) Mandarin. The symbol (>) is used here to indicate that the PH use of a bare wh-word seems to be grammatically more stable in Japanese than in Mandarin, but this requires confirmation in future research. That is, unlike Mandarin, Japanese has a wh-word (i.e. nani 'what') that can be productively used as a PH without any morphosyntactic remedial operation, but it is not obvious (without a statistical analysis) whether the PH use of bare wh-words is more stable in Japanese or in Mandarin on the whole.

Implications and remaining issues
We have described the wh-derived PHs in Japanese, Korean, and Mandarin. Our central claims are: (i) the PH use of a bare wh-word may be grammatically unstable, and (ii) in such cases, it requires a certain morpho-syntactic support. We have also argued that cross-language and intra-language variations are observed in connection with each of these two claims.
As for claim (i), the degree of grammatical stability varies from language to language (e.g. Korean bare wh-words may act as a PH in many cases, whilst Mandarin wh-words usually need to combine with na 'that'.), as well as within a language (e.g. Japanese nani 'what' may behave as a PH productively, but this is not the case with the other wh-words.). grammatical function (i.e. the place-holding function). Also, one might argue that the combination of a wh-word and a demonstrative is a case of lexicalisation. Such claims, however, must be supported by historical evidence. Since we can neither offer nor discuss such evidence here, the issues of language change must be left for future work. 20 Another avenue for future research is to extend our enquiry to other languages. As stated in Section 2.3, wh-words are a common source of PHs, but there are many other derivational sources (e.g. demonstratives, light nouns, indefinite pronouns). It would be a rather remarkable fact that all three East Asian languages surveyed here feature wh-derived PHs and that they display a similar set of morpho-syntactic remedial operations (e.g. doubling). As suggested by an anonymous reviewer, it would be interesting to find out whether these affinities are accidental or have some areal motivation. In this connection, we note that Cantonese, another East Asian language, exhibits doubling of a wh-word, as illustrated in (37).

(37)
Aaming houci sik-zo matje∼matje. Aaming seem eat-PRF what∼what 'Aaming seems to have eaten something (with the implication that the thing that Aaming seems to have eaten was previously mentioned to the speaker).' (Lee and Wong 2018: 339) [Cantonese] As the translation shows, however, the doubled wh-word in (37) does not function as a PH but as a 'reportative indefinite' (cf. Koev 2016). We must wait for another occasion to investigate whether grammatically unstable PHs are observed in languages besides Japanese, Korean, and Mandarin and, if the answer is positive, what morpho-syntactic operations are available to enhance the acceptability of such unstable PH forms.