Федеральное государственное автономное образовательное

учреждение высшего образования

Национальный исследовательский университет

«Высшая школа экономики»

3.1 Research methods

In the current research, I use the following methods.

Typological method of data collection from a language sample. The main sources of typological data, applicable for the current research, are language descriptions, grammar sketches or field questionnaires. As far as our target is to define world languages' trends toward indirect question marking, it is optimal to use a language Diversity value sampling method [Bakker & Miestamo 2016: 244]. In typological research, samples are constructed in such a way as to contain a representation of the most diverse languages to catch as many target patterns as possible. Normally, the diversity parameter is defined with the genetic and areal features of a language. A sample may be extended, for example, in case an abnormal diversity is captured among languages within a single areal or language family. However, in the current research, we are additionally restricted by the need for comprehensive descriptions, since not all grammars contain information on the indirect questions (so-called “bibliographic bias” [Bakker 2010]).

Corpus data processing method. Concerning the fact that documented grammar overview of indirect questions is frequently rather poor, additional sources of language material are useful. The method lies in corpus data analysis.

Syntactic typology method. The method aims to describe and observe collected data, specify terminology glossary to make data verifiable, to compile generalizations. Syntactic typology meets considerable challenges in view of the fact that key features of syntactic constructions cannot usually be analysed as segmental markers, e.g. like in morphology. Due to that, the comparison is kept to a restricted number of comparative concepts [Haspelmath 2016]. Also, syntactic typology method requires strict definitions of the concepts to make syntactic constructions comparison legitimate.

3.2 Description structure

For the purposes of the current research, I selected 30 languages as a sample. While describing data, a single indirect question construction is analyzed with the use of a number of comparative parameters, which allows me to inspect and compare languages with completely different syntactic structures. The analysis is presented in a database format in Google Sheets. If it is known that a language has more than one way to express indirect questions, they are all documented in a database.

A comparative parameter is a binary (yes or no) or categorical feature of construction. The features are elementary and can be extracted from minimal language material or taken for granted from the author's language description. All comparative parameters, available for construction, are:

Language info

For each construction basic language information is presented:

- language name;

- ethnologue code;

- genetic affiliation (language family name and first group level);

- geographic info (where is a language spoken);

- language area.

Construction info

There is some basic information on construction type:

- schematic construction description;

E.g. V=ni construction for Lezgian language, where =ni stands for interrogative clitic attached to the verb stem [Haspelmath 1993: 417].

- question types, encoded by the construction;

There are three possible categories for question types: “polar”, “wh-question” and “alternative”, that can be combined in any way. E.g. if a construction is used for expressing indirect polar and indirect specific questions, it will be marked as “polar & wh-questions”. Independent question types, if they are supported with construction, are ignored in the current field.

- diachrony data in free text format, if available, it is not used as a comparative parameter;

Also, construction is analysed with a range of binary features. Features listed below include information about means of indirect question encoding. A feature receives a positive truth-value in case it persists in a construction. More that one feature may receive positive truth-value for a single construction.

- question word/particle;

In interrogative sentences constructions make use of special question particles and question words. Some languages use in embedded constituent questions the same set of question words as in independent interrogatives, and it is the only indirect question marker in construction (e.g. Russian, Chuvash, Ket). In other languages, an additional interrogative particle or clitic is also needed (Udihe, Mian, Tundra Nenets languages). In polar question, particles may be used standalone (English, Adyghe). Such particles may be either specific for embedded questions or the same as in independent interrogatives.

- subordinator;

In some cases, a question particle may occupy a complementizer position, for example, English “whether”. Also, indirect questions may be expressed with the same complementizer that is used in regular complement clauses as well (English “if”). A special non-finite verbal form is also considered as subordinator (e.g. nominalization, participle, etc.).

Following [Cristofaro 2003] I distinguish three types of subordinators: complementizers, adverbializers and relativizators. These three types are defined in synchronous polysemy section and also considered in data overview.

- quotative;

Indirect questions may be marked as quotation construction with the use of quotative markers. Such quotations are expected to preserve the properties of independent interrogative sentences.

- disjunctive particle;

Disjunctive particles may mark embedded question constructions. Such constructions are rather rare and used in polar question marking. Cf. (7) from Supyire (Atlantic-Congo > Niger-Congo):

(7) Kа uru nа?i <...> mмм yнgй na uru щ

and this(EMPH) man.DEF <...> me ask that he(EMPH) he

? sб mмм yаha mobнlнge e lаa.

SUBJUNC go me leave truck.DEF in or

`Then that man <...> asked me if he should go take (lit. leave) me in the truck.' [Carlson 2011: 454]

- specific verbal construction;

Specific verbal construction field includes a set of constructions including more than one verbal form. One of these includes the disjunctive-negative constructions introduces in [Kцnig & Siemund 2007]. According to the definition given in [Kцnig & Siemund 2007: 297], a disjunctive-negative (“V not V”) construction involves an alternative question used to encode polar questions (lit. Peter is coming or is not coming?).

In some lgs embedded questions can be introduced by specific constructions including a set of devices such as::

Ѓ negation (a negative verb form or a special negative lexeme);

Ѓ duplication of full or reduplicated verbal phrase or negative word form corresponding to verbal phrase;

Ѓ duplication of question marker on both verbal/negative word forms (except for “null” disjunctive-negative variant without any question markers at all;

Ѓ absence of disjunction between verbal forms (although disjunctive particle may be used as indirect question marker, so, in such case, it is to be attached to both verbal forms) -- considerable point, that syntactically differs alternative questions from special verbal constructions.

For instance, the Chuvash (Oghur > Turkic) example given in (8) looks like alternative interrogative in which disjunction is missing, and a second alternative is a negative form of the first one:

(8) vas?? x?j u??l-ma kaj-n-i-be kaj-m-an-in-e

Vasya self walk-inf go-pc_pst-p_3-ins go-neg-pc_res-p_3-obj

pлl-mes

know-neg.npst[3sg]

`Vasya doesn't know whether he will go for a walk'. [Egorova 2020]

In Mandarin Chinese (Sinitic > Sino-Tibetan), there is no special question marker (so-called “null” strategy), but a head verbal phrase is duplicated in a negative form:

(9) ni mбng bщ mбng?

you busy not busy

`Are you busy?'

So, according to the logic of the current database, in case one construction may be subdivided into components, that may be used as independent indirect question constructions, they are presented in a database as separate units. For example, in Chuvash (Oghur > Turkic), there is an interrogative clitic =?i, that may be used in any type of indirect question. Also, it is used as a part of special verbal construction [Egorova 2020]. So, in the database, we have regular V=?i and special verbal construction V=?i Vneg=?i as different constructions.

It is possible that no one of the listed features will be marked with the truth value, e.g. in case construction is specified with intonation pattern or with a special verbal form. Also, more than one value may be true for construction, e.g. in case that is specific verbal construction with the use of question particle (like in Chuvash, mentioned above). In further discussion, I prefer to call such a combination of features as specific complex constructions.

Question types info

In the current research, I observe question types listed below, following definition given in [Kцnig & Siemund 2017]:

- polar questions;

- constituent questions;

- alternative questions;

Alternative questions are subdivided into two types: polar alternative and constituent alternative. The polar one expects truth value as an answer (cf. 10a), with the other one constituent answer is expected (cf. 10b):

(10a) Do you love me or not?

(10b) Do you want tea or coffee?

Also, I observe the affection of focus on polar and constituent questions' construction. Following [Krifka 2013: 1742-1785], a question may have more than one possible answer regardless of question type. Expected one is indicated with the use of focus or its absence. Cf. (11a), (11b) for a polar question and (11c), (11d) for a constituent question [Krifka 2013: 1751, 1780].

(11a) Did BILLF go to the party?

(11b) What did BILLF bring?

(11c) What happened?

(11d) What will happen tomorrow?

I assume that focus may affect syntactic features of indirect question constructions. In view of that, I also add information on focussed polar and constituent questions if possible.

Such a question type profile is filled for both direct and indirect interrogatives in order to define whether construction is specific for indirect questions or not.

Additionally, for indirect questions profile section I have a sub-division for the semantic type of matrix clause predicate. I observe two semantic groups: speech and mental verbs as it was attested, that for some languages indirect question marking may be contrasted for them. For example, in Chuvash (Oghur > Turkic), `know'-type verbs require a participle strategy in regular complementation, and interrogative complements may be also marked that way. However, that is impossible for speech verbs [Egorova 2020].

Synchronous polysemy of marker

In some languages the embedded question markers are polysemous with other constructions, e.g. some types of subordinate clauses. For example, as it was mentioned above, Chuvash (Oghur > Turkic) uses its main complementation strategy to encode embedded questions. Also, languages frequently use conditional markers as subordinators in indirect interrogatives (e.g. English if). The Humburi Senni (Songhay > Nilo-Saharan) shares the same marker with both relative construction and indirect interrogatives.

In the scope of current research, I document such polysemy evidence and attempt to compile implications.

Word order

Word order may be the only feature that opposes direct and indirect interrogatives (for example, in English and German indirect constituent question). I observe the feature as a binary feature and document, whether the word order in indirect question marker differs from direct interrogatives, in case the latter one is possible for construction.

4. Results

The current overview is an analysis of 30 languages from 20 language families and 2 isolates. There are 78 embedded question construction entries (see Table 1). In order to select language material, 95 grammars, articles and field trip reports were observed. Only 36 of them contained some information about indirect question constructions.

Polar and constituent questions are described in detail in many grammars. However, grammars do not provide enough material for typological conclusions on markers' diachrony, so, it will be left out of the scope of the current discussion. Also, indirect alternative question construction does not suffice for general overview as it is not covered in grammars. Given such an under-researched topic, the sample size does not allow to draw any generalizations. Thus, I will concentrate only on rare features of alternative questions.

Table 1. Language data

As for strategies that are defined in section 3.2.2 as binary features, we have the following distribution:

Table 3. Indirect question strategies:

In the current section, I draw attention to each strategy mentioned in table 3 (subsections 4.1-4.6), then the other parameters affecting indirect question constructions will be observed (subsections 4.7-4.10). In section 4.11, I discuss some rare types that are insufficient for typological conclusions but surely need further investigations.

In all examples presented in the current section the authors' glossing, transcription and translation are saved.

4.1 Question words and particles

First of all, I want to separate question word constructions and question particles constructions. Question words (interrogative pronouns), often called wh-words, are used to ask direct and indirect constituent questions. I define question word constructions as “null” strategy because usually there is no syntactic difference between embedded and independent question constructions. See details in section 4.6. syntax polysemy verb sentence

A question particle is usually a cliticalized element, that may be attached to a definite phrase. Cf. (?) for Mian (Ok > Trans-New Guinea) language, where a constituent question particle =e is attached to a dependent verb phrase:

(12) naka=i wan-оta tl-aa-ib=e

man-PL.AN who-PL.AN.EMPH come.PFV-DEONT-2/3PL.AN.SBJ=CQ

ga-b-e=ta

say.IPFV-DS.SIM-3SG.M.SBJ=MED

`...and was thinking who should come' [Fedden 2011: 529]

In Chuvash (Oghur > Turkic), interrogative clitic =i/j may be attached to any constituent, depending on information structure [Egorova 2020] (see details in section 4.8). In Momu (Fas) embedded polar question (13), interrogative clitic =fa is added to a clause [Honeyman 2016: 350]:

(13) Ay yeni-f=fa “Flerwick ai-ta=fa?”

2sg say-2sgS=ynq Flerwick anim:there.be[1|3sgS]-stvzr=ynq

`Did you say (ask) “Is Flerwick around?”' [Honeyman 2016: 559]

Question particle strategy is rather frequent in my data. In general, there are 19 entries of such strategies, that are used in the following way:

Table 2. Question particle strategy types

According to the data given in Table 2 it can be seen, that question particles tend to be specific to question types. That may be explained by the fact that languages usually combine different types of indirect question constructions, so question particles may be used for one type of questions, while the other one does not include them at all. Only 3 out of 13 languages containing question particle constructions have specific particles for constituent and polar questions. Otherwise, the only particle is used for all question types (2 languages) or question particle in one question type is combined with completely different constructions (e.g. subordinate or disjunctive) in the other one (8 languages).

As it was mentioned above, the information on means of alternative question is left uncovered in most of the grammars. In view of that, I have not included alternative question data in the overview above in combination with other types of questions. The only evidence, when a particle marks alternative language standalone is discussed in section 4.11 as a rare feature.

For instance, in Jamsay (Dogon > Niger-Congo), the question particle ma is used in all the types of embedded questions. Cf. (14) for a polar question and (15) for constituent question. In the case of constituent interrogatives, question particles are used alongside with wh-words:

(14) [ъ y?Мr??: mа] j?м ъjъrу-sа-bа

[2Sg come.Impf Q] say ask-Reslt-3PlS

`They asked whether you are/were coming.' [Heath 2008: 476]

(15) [a:=y nм-dо:n w?М mа] j?М:-gу-m

[who?=Foc here be.Hum Q] know-ImpfNeg-1SgS

`I don't know who [focus] lives here.' [Heath 2008: 477]

In Ket (Northern Yeniseian > Denй-Yeniseian), the question particle ы is used in indirect polar question (16). In indirect constituent question, “null” wh-word strategy is used.

(16) bы mana ??tn ы

3SG she.said 1PL QUEST

{i}8-a?a7-k5-s4-aq0-an-1

38-to.forest7-TH5-NPST4-go.MOM0-AN.PL-1

`She asked if we are going to the forest.' [Nefedov 2015: 62]

In some languages, question particles are specific only for embedded interrogatives, i.e. they cannot be used in root questions. Such feature is rather frequent: 5 constructions (2 of them are specific for wh-questions and 3 -- for polar ones) out of 19 may be used in indirect questions only. The other ones may be used in root question in any way. E.g. in Supyire (Atlantic-Congo > Niger-Congo), the constituent interrogatives' particle y? may be used in both independent (17) and embedded (18) questions:

(17) Dм fan?kй mаha n-tuga а jwu y??

how grave.DEF HAB IP-dig SC say Q

`How is the grave dug?' [Carlson 2011]

(18) U a cи jи х sн mМ-pа y?.

he PERF know who s/he FUT FP-come Q

`He knows who will come.' [Carlson 2011: 455]

By contrast, Basque (Isolate) complex construction ba al V-n containing question particle al cannot be used in independent questions (19).

(19) Erregeak zerbait egiten ba al zekien

king.ERG something do.lMPF ba Q knew.COMP

galdetu zion.

ask AUX

`The king asked him whether he knew [how] to do anything.' [Hualde 2011: 483]

According to given data, types of direct and indirect questions that can be marked with a given particle, always correspond to each other in case a particle allows marking of both direct and indirect questions. E.g. in case a particle is appropriate only for indirect polar questions, then it is possible in direct polar questions (e.g. Chuvash, Ket, Bardi). Indirect constituent question particles as well may be used in direct wh-questions (e.g. Momu, Supyire).

4.2 Subordinators

Indirect question marking with the use of subordinators is the most frequent strategy in the given data. There are 31 instances of subordinators strategy out of 78 entries in the database.

Firstly, subordination may be expressed as a special non-finite verbal form, e.g. in Alaskan Yupik (Eskimo > Eskimo-Aleut) (20), uci?-nominalization is used for embedded polar and constituent question encoding:

(20) Nallu-aqa [arna-m neq-mek

not.know-IND.1sg.3sg. woman-REL.sg. fish-ABM.sg.

nere-llru-ci-a ataku-mi]

eat-PST-VNnm-ABS.3sg.sg. evening-LOC

`I don't know whether the woman ate fish in the evening.' [Miyaoka 2015: 172]

Secondly, a subordinating particle or clitic may serve as a component of indirect question construction. For instance, in Hindi (Indo-Iranian > Indo-European) (21), conditional particle ki is found as an embedded interrogative marker. Also, it can be noticed, those subordinators may occur in constructions with wh-words, so, subordinators strategy is possible for constituent questions:

(21) u:љa:ne pu:cha: ki m?Ю kyх ga:u: ja:u:ga:?

Usha-erg asked that I why village go-fut

`Usha asked, why should I go to the village?' [Koul 2009: 182]

Third, a construction may be a complex of subordinating elements: e.g. in Basque (Isolate), an embedded polar question may be introduced with a construction ea ba-V-n. Ea is a conjunction particle, ba is a conditional marker, and -n is a relativizer:

(22) Bidriosek ea bazuen Institutuko berririk galdetu

Bidrios.ERG ea ba.AUX school.REL new.PRTT ask

zion.

AUX

`Bidrios asked him whether he had any news from the school.' [Hualde 2011: 483]

Subordinator strategy is distributed across question types as it is presented in the following table:

Table 3. Subordinator strategy question types

So, it can be seen that subordinators, unlike question particles, tend to be unspecific for various question types. In addition, it is a rare situation, when a language uses two subordinators in order to encode different question types. In common, a combination of subordinator with another indirect question strategy is preferred. However, there some languages, where two different subordinators are actually used that way: e.g. in Humburi Senni (Songhay > Nilo-Saharan) there is an embedded polar subordinator (23) yeмlaМ (analogous to English (Germanic > Indo-European) indirect question marker whether) is used along with an indirect constituent subordinators (24) kaм or gaм, and none of them may be used in another question type:

(23) iМ siм: ?beмy [yeмlaМ aМ? kaМ]

1SgS ProgrNeg know [whether 3SgS come]

`I didn't know whether (or not) he/she had come.' [Heath 2014: 366]

(24) iМ suМ ?beмy [boМr kaм kaМ]

1SgS ImpfNeg know [person Rel come]

`I don't know who came.' [Heath 2014: 367]

As for synchronous polysemy, markers are distributed across subordinate relation types as it is shown below (for more information in synchronous polysemy see section 4.10):

Table 4. Synchronous polysemy of subordinator

Specific subordinators used inr indirect question constructions without any evidence of polysemy are quite rare. In current data there are only two instances of such a situation, e.g. English (Germanic > Indo-European) indirect question marker whether seems to be impossible outside embedded interrogatives.

It is possible, that a subordinator may serve in more that one relation type: for example, in Bardi (Nyulnyulan), gerund prefix ma-, which serves as embedded polar a constituent question marker, also possible in complement (25) and adverbial (26) relations of purpose:

(25) I-ny-jargi-gal ma-yoorndoorroo-n-ngan iidarr.

3-PST-fear-REC.PST GER-cross-CONT-ALL creek

`He was afraid to cross the creek.' [Bowern 2013: 671]

(26) Iina noorro-ngan ma-loorroo-n.

firestick fire-ALL GER-light-CONT

`Iina (firesticks) are for lighting a fire. [Bowern 2013: 671]

Moreover, in Adyghe (Circassian > Northwest Caucasian) z?-V-case indirect question construction may be used in adverbial clauses as well as in relative ones.

Complement relation is the most frequent type of synchronous polysemy. It is expectable because an indirect question is commonly a complement clause of mental or speech verbs. So, it is unsurprisingly, that a language may not have a specific strategy. Then, adverbial polysemy is rather frequent is rather spread as well. Following types of adverbial clauses are attested:

Table 5. Adverbial clause types

So, as it was expected, conditional particles are frequently used in indirect question constructions. E.g. English (Germanic > Indo-European) if is used either in both embedded polar question and conditional clauses, cf. (27) and (28):

(27) Let me know if there is anything I can do. [Eckardt 2007]

(28) If I were you, I'd run away. [SketchEngine]

Also, the concessive meaning is usually marked the same way as conditionals, e.g. Lezgian (Lezgic > Northeast Caucasian) -t'a (29):

(29) Hik' иьnьx-na-t'a sadra sьhbet aja kwan.

[how steal-AOR-CND] PT talk do:IMPV PT

`Say how it got stolen.' [Haspelmath 1993: 426]

As for relative relation, it is rather rare in the given data. Interestingly, some language may use relativization in order to produce a constituent question. Consider an example from Hubmuri Senni (Songhay > Nilo-Saharan) example (24) above. Questioned constituent it replaced with indefinite noun phrase boМr `person' with a relative adjunct, so literally the example may be translated as `I don't know the person who came'.

Subordinators sometimes may be used in independent sentences, but it implicates special semantics (e.g. rhetorical, hesitation). For instance, in Tundra Nenets (Samoyedic > Uralic) conditional marker may be used in a head verb (30). Such question invokes rhetorical semantics:

(30) ??c?ekex°yu-naq t?d° n?er° n?ah yil?ebc??-d?ih x?c?er°

child.DU-1PL then forth to life-3DU how

??-?ko-b°-ta=w°h

be-FUT-COND-3SG-DUB

`How on earth will the future life of our children be?' [Nikolaeva 2014: 58]

However, the information on such types of questions is rarely covered in grammars, so, the data is insufficient for typological conclusions.

4.3 Quotatives

The quotation is rather frequently used as a mean of indirect question encoding. However, the quotative strategy meets a challenge as sometimes it is used in direct quotations. The direct quotation is frequently used for indirect question marking and seems to be possible in the majority of languages, cf. English (Germanic > Indo-European) example (31) without any quotative and Rutul (Lezgic > Northeast Caucasian) one with quotative xur (32). In Rutul example, there is no clause embedding: there is just interrogation accompanied by quotatives, which is a rather frequent strategy, too.

(31) He asked me: “Do you want some tea?”

(32) hada xur=xa mi-d љuw xur=xa ha edemi-je-ra

PTCL QUOT=ADD this-ATTR what QUOT=ADD that man-OBL-ERG mi x?d?l-now-a, ч?n?ч чaи'i<r>ч?-r we hile-?

this woman-OBL.H-ERG child <1>throw.PFV-CVB you.PL where-LAT

du-ru?u-r=a?

HPL-go.IPFV-CVB=be

`They say: “Hey, that man, that woman, she's dumped her kid, where are you going?”' [Kina 2018]

Also cf. Mongsen Ao (Ao > Sino-Tibetan) example (33), where there is a clause embedding with a quotative complementizing particle t?М, but embedded interrogative is a direct quotation:

(33) nм n? ajim?щ nн t?М sa

1SG AGT cry.out-IMM QPTCL thus say.PST

`“Will I make a proclamation?”, said [Yellow-backed Sunbird].' [Coupe 2008: 149]

So, there are several possible patterns of quotation strategies in indirect question constructions that may be distinguished from the current data:

- direct question + quotative

- matrix clause + root question

- matrix clause + root question + quotative

- matrix clause + indirect question + quotative

In the current research, I prefer to leave any direct quotes constructions out of the scope of observation. On the one hand, first, direct quotation seems to be possible in every language, but usually not covered in grammars as it considered as conversational and low-styled. Second, a language usually has a huge amount of diverse conversational constructions seem, so, its documentation is to overload the database. Finally, direct quotation seems to have a range of universal patterns in languages, that are not comparable with indirect question constructions, because embedded interrogatives do not possess features of direct speech. E.g. deixis shift is peculiar for indirect questions and report quotation, but impossible in direct quotation.

On the other hand, there are some languages, where the direct quotation is the only strategy of indirect question encoding. Thus, such languages cannot be observed within the current database. So, the issue of direct quotation remains disputable.

Then, if we take into account only the latter pattern (matrix clause + indirect question + quotative), there is only two such entry in our database, therefore, it seems to be typologically rare. For instance, in Mian language (Ok > Trans-New Guinea) (34), a combination of question particle =a and quotative clitic =ba is used for indirect polar question encoding:

(34) [kwйit=e hei-€t'-ne

sugar_cane=SG.N1 cut.PFV-give.PFV-1SG.R

dowфn'-nab-e=a=ba?] ge

eat.PFV-NRPST-3SG.M.SBJ=Q=QUOT say.PFV

baa-n-o=ta

say.PFV-SEQ-3SG.F.SBJ=MED

`“Did he cut and eat my sugar cane a short while ago?” she asked and then…' [Fedden 2011: 491]

In Chuvash (Oghur > Turkic), quotative is possible (35) with both speech and mental verbs in various question types:

(35) va?? ?t-r?-? mлn bol-za te-ze

Vasya ask-PST-3SG what was-CV_SIM QUOT-CV_SIM

`Vasya asked what has happened.' [Egorova 2020]

In that case, there is an evidence of the possibility of such construction with non-speech verbs, which is quite unexpectable. However, the data is insufficient for typological conclusions.

4.4 Disjunctive markers

In the strategy, a disjunctive marker or particle is used as a subordinator in indirect question constructions. The disjunctive strategy is rather rare in a given data, there are 5 entries out of 78 constructions. For example in Tamashek (Berber > Afro-Asiatic) disjunction wжlб `or' serves as a question marker in embedded polar questions with the mental verbs in the main clause (36).