In both descriptive and comparative linguistics, the study of the level of discourse and “pragmatics” has long been on the periphery. Multiple grammatical descriptions of particular languages containing fine-grained analysis of morphology and syntax are sometimes strikingly incomplete in the parts devoted to what is usually labeled as “particles”, “pragmatic markers”, “discourse particles”, “discourse markers”, etc.

In typology, the advanced cross-linguistic study of discourse-associated elements can be traced back to the 1980s, an important milestone being the edited volume by Weidt (1989). Not surprisingly, the studies of discourse particles etc. evolved in close association with the linguistics of German: the extensive use of so-called “modal particles” (Modalpartikeln, Abtönungspartikeln) is a hallmark of this language.

The objective of this paper is twofold. First, I provide a working typologically-informed framework for analyzing the syntax and functions of the elements of Lithuanian which are normally termed “particles” in the literature. Within this framework, I focus on two “particles” of contemporary Lithuanian, namely juk and gi, whose functions and morphosyntactic² properties have never been given sufficient attention. Juk and gi are considered together, as they exhibit a certain degree of functional overlap, although there are also important differences. Giving an account of the functioning of juk and gi is my second and main goal. In doing this, I invoke cross-linguistic comparison, namely, the studies of German particles doch and ja. The latter exhibit interesting functional parallels with juk and gi but are much better described in the literature.

Although this study is primarily descriptive, it also appeals to the typological community, and addresses the issue of commensurability between language-particular structures. The reader can find some discussion of the relevant methodological debate in 2.1 and 4.

In Section 2.1, I address the issue of terminology in the studies of discourse-associated elements. Section 2.2 discusses the morphosyntactic classes of “particles” in Lithuanian, and a new classification of this kind of elements is proposed. In 2.3, I take a closer look at the positioning of juk and gi. Section 3.1 discusses the morphosyntax and the functions of the particles doch and ja in German. In 3.2, I analyze the functional contribution of juk and gi in Lithuanian. In 3.3, the particles of German are compared with the Lithuanian ones. In 4.1, I highlight the contribution of the present study to the current methodological debate on “comparative concepts” (Haspelmath 2010) in linguistic typology. In 4.2, the areal-typological and diachronic dimension of the functions exhibited by juk and gi is briefly introduced. In the Conclusion, I summarize the findings of the study.

In the last decade, there has been some debate on the terms used for markers exhibiting functions usually ascribed to the domain of discourse, both cross-linguistically and language-specifically.

Both aspects of the problem – the domain of discourse itself and the labeling of markers associated with it – are problematic. First, it is hardly possible to draw a clear boundary between “discourse” and “non-discourse” (e.g. grammatical) functions. In this paper, I take an agnostic position and do not attribute the functions studied here to “discourse” or any other domain. When describing the functioning of juk and gi in Lithuanian, this would not contribute any additional information.

The most popular labels for the markers associated with the domain of discourse, regardless of what authors have in mind, are “discourse markers”, “modal particles”, “discourse particles”, “pragmatic particles”, and many others. There are numerous works which discuss the differences between these labels, the most influential recent crosslinguistic studies (in non-generative linguistics) being Degand et al. (2013) and Fedriani & Sansò (2018). However, the problem with both volumes as well as many other works is the aprioristic approach, which has been extensively criticized by Haspelmath (2007, 2010, 2018). Although this stance is not always made explicit, authors believe that there are natural-kind-like cross-linguistic categories such as “discourse markers” or “modal particles” that manifest themselves in various ways in different languages, and one can “diagnose” them attributing language-particular elements to one or another crosslinguistic kind. For this paper, I refuse such an approach and take a non-aprioristic stance in this respect, accepting the view that morphosyntactic criteria are language-particular.

This does not mean, however, that the categories of various languages are completely different and incommensurable. On the contrary, the morphosyntactic categories of different languages sometimes are strikingly similar, especially when the languages are genealogically related or geographically adjacent. In such cases, it is indeed convenient to use a single term in order to talk about the categories of different languages, being aware that there are, however, differences between them, and highlighting these differences.

As many other language-particular descriptions, the most widely used comprehensive grammar of Lithuanian (Ambrazas 1997) contains a section entitled “particles” (Lith. dalelytės). This section is an overview of the elements with a very broad scale of morphosyntactic and semantic properties. In fact, what unites them is that these elements do not fall easily into other function-word classes such as conjunctions, connectives, or propositions. These elements are typically short (1-2 syllables). The criteria for semantic and syntactic classification of particles provided in the grammar are not made explicit, and sometimes the motives for grouping elements in one or another way remain obscure.

The elements we focus on in this paper are classified as “intensifying-emphatic particles” in Ambrazas (1997, 401–402). Besides juk and gi, the connecting element ir is also included in this group. All three “emphasize a word or a clause”. In my view, there are at least two problems with this definition. First, it is not made explicit what is understood under “emphasis”: this notion seems far too vague and permits a large range of interpretations. Second, modifying a “word” or a “clause” are two very different things in terms of semantic scope. In fact, in the examples of gi in which it is claimed to intensify a single word, the particle’s scope is the whole sentence (“The particle gi, usually postposed to the word it intensifies, sometimes occurs in the initial position” [Ambrazas 1997, 402]):

(1)

^{3 , 4}

In both examples, the particle occurs in direct contact with the verb following (1a) or preceding it (1b). In (1b), the first clause consists of a single verb predicate. If the scope of gi were the “word”, or a single constituent consisting of a single head word which is a verb in this case, one would suspect gi to be able to take another constituent as its scope if it is moved to a different position. If we take a slightly longer sentence, one can easily observe that this is not the case.

(2)

Ex. (2a) is the most natural and strictly preferred in contemporary Lithuanian, (2b) and (2c) are reported possible in spontaneous speech (2b is better than 2c), (2d) is highly unlikely (possible only in a self-correction), and (2e), in which gi follows the first direct object, is impossible. In (2c) gi, although it follows the second direct object, this constituent is not focused: it is still the whole sentence that the particle takes scope over, as in (2a) and (2b). One can conclude that the scope of gi is the hole sentence regardless of its position,⁵ which is most often the second one (“Wackernagel”), but sometimes also the sentence-initial (1b), the postverbal (2b), or the sentence-final one (2c).⁶

Turning back to (1b), we are dealing with a biclausal construction in which the main clause consists of a single verb. Here, gi modifies the whole construction. This can be observed if one, e.g., extends the first clause (3) or moves gi to the end (4).

(3)

(4)

The same observation holds for juk, the main difference being the preferred position which is sentence-initial in the latter case.

Unlike gi and juk, the elements such as ir ‘also’ and net ‘even’ (included in the same “intensifying-emphatic” category in Ambrazas, 2006) exhibit a different scope-related behavior. For instance, the example with ir taken from the grammar can be modified in the way shown in (5):

(5)

From the example (5), one sees a clear difference in meaning when ir occurs with another noun phrase, and it is the phrase and not the sentence which is in the particle’s scope. Ir may have scope over words and phrases of different kinds; consider the various readings of the same sentence when ir has different scopes (6):

(6)

The same applies to net ‘even’, which also has a narrow scope, and the interpretation of a sentence depends on the scope of ir or net.

Concerning the analysis above, I propose a set of criteria which are relevant when describing the elements of Lithuanian traditionally termed “particles”; such criteria allow us to adequately reflect their morphosyntactic properties.

1. – scope: sentence vs phrase. I propose to call the Lithuanian particles with sentence scope sentence particles⁷ regardless of their preferred position in the sentence, following in part Munaro & Poletto (2004).⁸ Unlike sentence particles, phrase particles take scope over a single constituent within their clause.⁹

2. – position. This criterion defines the preferred or obligatory position in the sentence for sentence particles and the position with respect to the constituent under the scope for phrase particles.

3. – boundness: ability vs inability of an element to occur in isolation. This is Bloomfield’s (1933, 160) criterion which reveals the degree of morphosyntactic autonomy of an element in a language. One can also cite Haspelmath’s extended formulation: free forms can form a complete (possibly elliptical) utterance (Haspelmath 2013, 213). Thus, the “particle” elements of Lithuanian can be classified as either bound or free.

4. – prosodic dependency: Integration into the intonation contour of an utterance. This criterion applies only to sentence particles. Elements can be classified as either integrated into their host sentence’s intonation or non-integrated (having their own intonation pattern).

Table 1 presents some frequently-used elements classified according to the values of the listed parameters apart from the one of prosodic dependency.¹⁰

Table 1
Morphosyntactic types of “particles” in Lithuanian

The list presented in Table 1 is not exhaustive and only serves as an illustration of the application of some of the listed criteria. However, my suggestion is to conduct future research on Lithuanian “particles” having in mind these classification principles, avoiding the confusion of elements having very different properties.

I will now focus on the morphosyntax and functioning of two elements of contemporary urban spoken Lithuanian – juk and gi. Both elements fall into the type “sentence particles” if the proposed criteria are applied. Both take scope over the whole sentence, both cannot occur in isolation (are bound), both have preferred, although different, positions in their host sentences. Another and probably the single other element of Lithuanian which shares the same set of morphosyntactic properties is the polar question marker ar (7b), unlike its cognate and diachronic source ar which is a marker of alternative (7a):

(7)

Corpus data¹¹ indicates that gi strongly prefers the second position (97 of 100 random occurrences in the colloquial subcorpus of LKT), whereas juk occurs in the first position in (74 of 100). The notion “second position” here is to be interpreted as the position after the first full phrase in a sentence.¹² For example, (8a) is preferred to (8b) which sounds awkward if acceptable at all (in the case if mano ‘my’ is not focused and highlighted by the intonation):

(8)

¹³

The rare exceptions to the second-position rule for gi can be divided into two groups. First, gi sometimes starts a sentence. Out of 100 random uses in LKT (colloquial subcorpus), it occurs 3 times in the initial position.¹⁴ An example from the corpus is:

(9)

Other cases when gi is seemingly found in a position other than the second one may be described using the insights gained in Andrej Zaliznjak’s works on the second-position clitics in Old Russian. As Zaliznjak demonstrates (2008, 47–56), Old Russian clitics are able to “move” rightward from the second position in the cases when obligatorily sentence-beginning elements such as conjunctions or fronted constituent(s) (i.e. a topicalized or a focalized phrase) are present. In such cases, a “barrier” (Rus. ritmikosintaksičeskij barjer) is placed after these elements, and the second position is counted starting from the barrier, e.g.

(10)

Similar forces seem to be involved in Lithuanian. Gi also “moves” rightward in the sentence if some specific kind of element occurs in the beginning of a sentence. It may be a conjunction or a sentence-initial discourse marker as in (11a), or a fronted topicalized or focalized constituent, e.g., an adverbial as in (11b):

(11)

Gi sometimes (rarely) occurs sentence-finally. This use is perceived by Lithuanian speakers as non-normative and markedly colloquial. Such cases are not reflected in LKT, even in its spoken sub-corpus:

(12)

Finally, in colloquial language gi is able to occur after a focused constituent preserving its sentence scope. In the following example, Kaune ‘in Kaunas’ is said with an intonation which marks the sentence focus:

(13)

Ar, which is also classified as a sentence particle here, exhibits a similar behavior: its neutral position is sentence-initial (14a), but in colloquial language it is able to occur preceding the focalized constituent if there is any (14b). In (14b), the scope of ar is still the whole sentence: the entire sentence preserves its question illocutionary force.

(14)

In both sentences, Kaunas is focalized. This is marked by the focus intonation with which Kaune is pronounced. The sentence turns unacceptable if one preserves the focus intonation on Kaune but moves ar to another constituent, e.g., before gyvena ‘lives’. Such a movement is only possible if gyvena becomes focalized, and this should be reflected in an intonation change (the focus intonation moves to gyvena)¹⁵.

The particle juk exhibits more positional flexibility. Although its preferred position is sentence-initial (74 of 100 random occurrences in the colloquial subcorpus of the LKT), it also frequently occurs in second position. It is also able to attach to focalized constituents, either preceding or following them.

A reason for comparing elements of different languages is the ability of these elements to translate each other. In the absence of an accessible German-Lithuanian parallel corpus, such a work cannot yet be done systematically.¹⁶ However, a preliminary study of existing bilingual texts shows that both gi and juk often correspond to both doch and ja in translations, cf. a fragment of the Lithuanian translation of Kafka’s Die Verwandlung (15 - 16):

(15)

(16)

¹⁷

I take this (at least occasional) “translatability” as the starting point and now turn to the German material.

German (as well as other West and North Germanic languages with the exception of English) exhibits a morphosyntactic class of elements which manifests properties similar to those I define as sentence particles in Lithuanian. In the German descriptive tradition, such elements are called Abtönungspartikeln (‘nuance particles’) or Modalpartikeln, and the labels modal particles or discourse particles are used in English-language publications. Modal particles constitute a clear-cut word class in German and some other West and North Germanic languages.¹⁸ These are characterized by two main morphosyntactic properties:

1. – modal particles are bound uninflected forms,¹⁹

2. – they occupy a strictly determined position in their host sentence (the so-called “middle field”, “Mittelfeld”): a slot directly following the finite verb (or sometimes the direct object) in declarative sentences and the subject in interrogative sentences.

Additionally, modal particles are typically unstressed and integrated into the sentence intonation. Another non-trivial and typologically non-obvious characteristic of the Germanic modal particles is the “form – function” correlation. Indeed, modal particles that are found in the middle field tend to express the attitude of the speaker toward the proposition (Bross 2012, 185), whereas “connectives” or “conjunctions” whose functional domain is relating the proposition to the previous discourse unit usually occur sentence-initially (but there are exceptions). All these properties resemble those of the Lithuanian sentence particles, the main difference being the strictness of the position in a sentence (rigid in Germanic, more loose in Lithuanian).²⁰

I will now focus on two German modal particles – ja and doch – whose functions, as I will show, are comparable to those of juk and gi. I now give a brief overview of how their functioning has been described in the literature.

From the very beginning, scholars noticed the functional affinity between ja and doch²¹. In fact, in many contexts they are interchangeable without a significant loss in sense. Consider an example in which the speaker starts a sentence with one particle, and then substitutes it with another after a pause in a self-repetition:

(17)

In earlier works, it was commonplace to claim that the core function of both doch and ja is to mark mutual knowledge or shared knowledge (Bross 2012, 197; Thurmair 1989, 104), or, formulated in a more formalized way, to relate a proposition to the common ground.²² The notion “common ground” is defined in the following way:

“[the mutual beliefs] of the parties to a conversation are the beliefs they share, and that they recognize that they share: a proposition Φ is common belief of a group of believers if and only if all in the group believe that Φ, all believe that all believe it, all believe that all believe that all believe it, etc.” (Stalnaker 2002, 704)

As Bross (2012, 197) argues, this notion is problematic: a speaker cannot really know what the interlocutor knows or believes as we cannot see into each other’s heads. In fact, both particles, especially doch, can be used when the hearer is clearly not aware of the proposition in advance:

(18)

In (18), the speaker believes that the hearer knew but forgot Peter’s virtues or even has an opposite opinion about him; the fact of Peter’s being suitable for the job position is beyond doubt for the speaker her/himself. Therefore, less strong formulations are needed, e.g. as the following one: “[the] feature that ja and doch have in common is that in both cases the proposition expressed – say, P – is taken for granted by the speaker, and is assumed to be taken for granted by the hearer as well” (Gast 2008, 187). A good label that captures this is the notion of uncontroversiality, more precisely, framing²³ the proposition as uncontroversial. This term is used to some extent by Rinas (2006), and it is crucial in Grosz (2010).

Although doch and ja share the functional core of framing the proposition as uncontroversial and are often interchangeable, there are also differences between them. The most visible difference is the presence of an adversative meaning component in doch and its absence in ja (Thirmair 1989, 110–119; Grosz 2010). Grosz (2010, 163) also formulates this meaning component as correction. This can be seen in example (19):

(19)

‘I am arriving.’ (Rinas 2006, 200)

In (19a), the speaker expects the hearer to be aware of her/his arrival. In (19b), the speaker implies that his/her arrival is contrary to the expectations of the hearer. The “correction” use may be seen as one of the manifestations of the contrastive component. In (20), the situation is described as follows: the hearer is amnesiac and forgot that s/he had been to Paris (e.g. says something like “I have never been to France”). The speaker corrects the hearer reminding him or her of the fact that this is not true:

(20)

The contrastive meaning component is most probably inherited from the source from which the modal particle doch evolved, namely the contrastive conjunction doch (a cognate of the English though) which is still in use in contemporary language, and occurs sentence-initially.²⁴

On the other hand, doch is unable to occur in contexts expressing general truth, something which is presented as obvious and known to everybody. In such cases, only ja is possible:

(21)

Another important aspect in which doch and ja exhibit differences is their ability to occur in various types of speech acts. In this respect, ja turns out to be more restricted and only occurs in declaratives, see Table 2:

Table 2
(Thurmair 1989, 44). Occurrence of ja and doch in different speech acts

On the basis of the analyses by Thurmair (1989), Karagjosova (2004), Rinas (2006), Gast (2008), and Grosz (2010), in Table 3 I provide a summary of the main contexts in which doch and ja occur (“functions”)²⁵. This is not exhaustive, but covers the most recurrent uses:

Table 3.
The main occurrence contexts of doch and ja in German²⁶

Having in mind the portrait of the German particles doch and ja sketched in the previous section, I will now turn back to the Lithuanian juk and gi. Their morphosyntactic characteristics have already been addressed in Section 2.2, and now I will discuss their contribution to meaning.

It is clear that many German sentences presented in the previous section can be translated into Lithuanian using either juk or gi, or both. Let us now consider some of their core uses.

As doch and ja, juk and gi seem to be interchangeable in many contexts. If one translates the German example (17) into Lithuanian, it is clear that both juk and gi are perfectly acceptable:

(22)

²⁷

The main difference between (22a) and (22b) reported by Lithuanian speakers is that (22a) sounds more formal, whereas (22b) is more colloquial. However, if one includes both sentences in a broader context, it turns out that juk and gi do not sound equally well. Imagine the following situation. The speaker is arguing that the social position of woman in the Roman society was relatively high, and this was related to the fact that Romans used to have only one wife. In this context, both (22a) and (22b) are equally acceptable. By contrast, if the interlocutor asks something like “How many wives did Caesar have?”, the variant with gi (22b) sounds better than (22a). With gi the speaker not only indicates that there is the uncontroversial fact that Romans used to have only one wife, but also that, in her/his view, there are inconsistencies in the hearer’s beliefs. This “contrastive” component of the meaning of gi is clearly parallel to that of the German doch (functions 2-3 from Table), whereas its absence is comparable to that of the German ja, although it seems that in Lithuanian this contrast is less strict: juk is still possible in contrastive contexts. The “correction” function of gi is even better observable in (23), which presents a fragment of a dialogue between a child and an adult:

(23)

²⁸

Speakers report that in (23) both gi and juk are possible, gi being, however, more natural, especially in colloquial speech.

The “reminding” function is illustrated in (24). The hearer seems to be surprised by the fact that it is early evening, and alcohol is not being sold in the supermarket. The speaker reminds her/him:

(24)

In (24), the speaker appeals to the supposed background knowledge of the hearer: alcohol is not sold in Lithuanian supermarkets after 3 p.m. on Sundays. In this example, gi is reported to sound better than juk, although juk is not completely excluded. Juk is more natural if the hearer is not expected to disagree and is supposed to be aware of both facts: (1) alcohol is not sold in Lithuanian supermarkets after 3 p.m. on Sundays, (2) today is Sunday. Imagine a situation in which two friends have a plan to have a beer on Sunday after 3 p.m. Theoretically, they have a choice: to buy beers in a supermarket and drink them at home or to go to a bar; in fact, only one option is realistic. The speaker A supposes the speaker B to be aware of both (1) and (2) and says:

(25)

In (25), both juk and gi are reported to sound equally natural (as in other cases, gi being more colloquial). The contrastive component is lacking here.

More visible and strict differences between juk and gi concern their ability to occur in non-declarative sentences. Only gi occurs in wh-questions and wh-exclamatives:

(26)

(27)

²⁹

In both cases, the use of juk would be impossible. However, juk can appear in questions if its scope is broader than the question itself, like (28) where the rhetorical question is “indented” within an intended declarative sentence:

(28)

In (28), juk takes scope over the intended proposition <It was he who gave them the monopoly rights> and not over the question only, which is rhetorical, and frames the whole intended proposition expressed by two sentences as uncontroversial.

It is not easy to explain how the marking of uncontroversial information is semantically related to the use of gi in questions. However, there are reasons to claim that gi only occurs in questions whose content is framed as expected by the speaker, consider example (29):

(29)

This sentence sounds natural if pronounced in the situation in which everybody is waiting for the results of the elections. By contrast, it can hardly be pronounced when the talk is about the weather, and, suddenly, someone has the idea to ask about the elections’ results. In the latter case, the sentence would sound more natural without gi (and would possibly start with beje ‘by the way’).

Both gi and juk occur in the questions equivalent with tag questions in English when the expected response is positive:

(30)

The combination of the polar question particle ar and gi (spelled as a single word argi) mark polar questions in cases when the speaker has doubts about the truth of the proposition under the scope of the question marker:

(31)

Finally, gi but not juk is used with imperatives in colloquial language. Such imperatives mark the impatience of the speaker who expects the hearer to be aware of the speaker’s wish but not to be acting for some reason as in (32):

(32)

In Table 4, I summarize the prominent contexts/functions in which the Lithuanian sentence particles juk and gi occur. As in the case of German (Table 3), these contexts are not exhaustive, though they do reflect the general picture and enable cross-linguistic comparison. (–) means that the use is allowed but dispreferred / not very natural.

Table 4.
The main occurrence contexts of gi and juk in Lithuanian

In the previous section, I analyzed the sentence particles juk and gi of Lithuanian using some insights gained in the studies of German modal particles doch and ja. I now provide a brief comparison of all four particles.

Concerning the positions in sentence of juk and gi, on the one hand, and doch and ja, on the other hand, these are defined on the basis of language-specific rules in both German and Lithuanian. These rules are different, and the German particles are characterized by stricter positioning (in the “middle field”).

The functions of all four particles are summarized below in Table 5.

Table 5.
Doch, ja, juk & gi: comparison

Indeed, all four particles of the two languages are characterized by the common meaning component which has been previously defined as framing the proposition as uncontroversial. German and Lithuanian exhibit two particles each, and in both languages there are contexts in which the particles are interchangeable. Moreover, in both German and Lithuanian, there are particles with broader and narrower functional scales (doch & gi vs ja & juk, respectively). In Lithuanian, as well as in German, the particle with a broader functional scale is characterized by the additional contrastive meaning component (doch and gi). For both particles, the presence of this component correlates with the ability to occur in non-declarative sentences.

Regarding the latter aspect, doch and gi also exhibit differences: gi occurs in non- declarative contexts more easily than doch. Doch does not occur in tag questions of the type 8. In such cases, a reformulation is needed (33a), or another particle, denn (33b), is used:

(33)

Moreover, unlike gi, the German doch occurs in wh-questions only marginally. The most typical occurrence of doch in wh-questions in German are the situations in which the speaker forgot something and wants the hearer to help recover the knowledge. It often co-occurs with another particle gleich:

(34)

In (34), the speaker expects her/himself to know who the person was who has just passed by, but suddenly s/he realizes s/he has forgotten the name.

In the cases like the Lithuanian (29), when the speaker asks a question with an expected topic, German normally uses denn:

(35)

To summarize, all the four particles of both languages exhibit a common functional core. Roughly speaking, the functional scale of doch is very much like the one of gi, whereas ja exhibits similarities with juk. In both cases, the coincidence is not complete, but there are more similarities than differences. It is also to be noted that the distribution of juk and gi in Lithuanian is freer than that of doch and ja: the Lithuanian particles are interchangeable in more contexts.

The methodological stance invoked in this study – using insights gained in descriptions of one language in order to understand structures and functions of another – refers to the current debate in linguistic typology known as the ‘comparative concepts’ vs ‘descriptive categories’ problem. The debate was triggered by Martin Haspelmath at the end of the previous decade, and is still ongoing. In Haspelmath’s (2010) view, particular languages should be described in their own terms (Boas 1911) in a structuralist/descriptivist spirit. Descriptive categories are always language-particular, and languages are best compared through comparative concepts consciously designed by linguists. The latter are clearly formulated, uncontroversial definitions of compared phenomena made up by linguists, and not pre-established categories or cross-linguistic categories, which, supposedly, do not exist as natural kinds (Haspelmath 2007, 2018). For now, one can say that Haspelmath’s main distinction between comparative concepts and descriptive categories has been generally accepted by the typology community.

However, this radical stance has also been criticized. Many points of criticism are collected in a special issue of Linguistic Typology (20(2), 2016). The main argument of most of Haspelmath’s critics is that comparison between languages and the knowledge of cross-linguistic patterns is doubtlessly useful in discovering language-particular structures. As Lander and Arkadiev (2016, 412) put it, “being informed about other languages and typological variation [...] in general enables one to [...] see systematic patterns in the data which otherwise might appear just chaotic.”

In this paper, I have followed the spirit of the study by van der Auwera & Sahoo (2015), which is one of the most influential papers criticizing Haspelmath’s approach. Arguing against his extreme categorical particularism (though accepting it in a softer version), the authors demonstrate that the distributional and functional properties of English such can be better understood if compared to similar elements of other languages. It turns out that the functional core of these elements becomes more visible in cross-linguistic comparison: all the languages mentioned in the paper exhibit elements with a non-trivial set of common properties, being more similar to each other than to anything else in their own languages. This enables the authors to propose a functional category (“similative”) which describes to a large extent the behavior of different language-particular elements.

In this study, I have demonstrated how an analysis originally designed for one language can be fruitfully applied to another language, very much as van der Auwera & Sahoo (2015) did. In doing so, however, it was important to keep in mind that descriptive categories of one language cannot be mechanically transferred onto another: commensurable language-particular categories always exhibit similarities as well as differences. These can be formalized in describing their functioning in the form of sets of elementary uses/ functions, as presented in Table 5. The data presented in Table 5 is a sort of comparative concept designed on the basis of only two languages – German and Lithuanian. On the other hand, this comparative concept gives one strong descriptive instruments for research on other languages.

German and Lithuanian are not located on different continents; both are Indo-European, both belong to the European macroarea (Haspelmath 2001), but also to the smaller area adjacent to the Baltic Sea (Koptjevskaja-Tamm and Wälchli 2001). They are historically interconnected, and their structures are definitely commensurable, although the degree of direct influence of one upon another is disputable. Although large-scale cross-linguistic studies of similar categories have not yet been published,³⁰ there are multiple language- particular descriptions in which some elements are described in very similar terms. Many such languages belong to the same Eastern, Central, and Northern European area. Obvious parallels to the particles discussed in the present paper are the Russian že and ved’ (see Valova (2016, 85–102) for an overview) and their parallels in Ukrainian (-ž, žež, adže) and Belarusian (ž[a]), the North Russian dialectal dak (Post 2006), the particles ju/ jo of the Scandinavian languages (Andvik 1994; Aijmer 1996), the -han/-hän “clitic” of Finnish (Palomäki 2009). There are some contrastive studies between pairs of languages: German vs Russian (Orlova 2012), German vs Dutch (Foolen 2006), German vs Czech (Rinas 2006). In the parallel Lithuanian-Latvian corpus, gi and juk regularly correspond to the Latvian taču.³¹ Beyond this part of Europe, similar elements are reported for the Northeast Caucasian (Nakh-Daghestanian) languages: Standard Lezgian (Haspelmath 1993, 242) and Sanzhi Dargwa (Forker, forthcoming)³².

Another aspect relevant for understanding the typology and areality of the elements framing proposition as uncontroversial is their diachrony. Regarding the particles considered in the present paper, the diachronic path is clearer for German than for Lithuanian. The German modal particle ja was already present in Old High German (spelled as ia) in a function close to the contemporary one (Wauchope 1991, 93–131), but it used to occur in positions other than the “middle field”. It had originated form the homophonous word meaning ‘yes’³³ which is reconstructed for Proto-Germanic (Kluge 1989[1883], 338) and survives in contemporary German. The particle thoh (>doch) also functioned in Old High German in the way resembling the contemporary one, and it derives from the homophonous adverb/ conjunction which has a contrastive meaning (‘though, however, but’)³⁴.

Both juk and gi are attested already in the earliest Lithuanian written texts, but their etymology remains somewhat obscure. Juk also occurs as jukaigi, jukaig, jukaigei, jukag, jukig in Old Lithuanian (Smoczyński 2017, 527), and its function seems identical to the contemporary one; consider an example from Daukša’s Catechism (1599):

(36)

The etymology of juk is indicated as obscure by Smoczyński (2017, 453). Ostrowski (2015) suggests deriving it from the combination of juo ‘especially, notably, all the more’ and kai ‘when’. The path of development of the particle gi is even more obscure. In Old Lithuanian, the particle clearly had a phrasal, not sentential scope, probably marking some kind of focus as in the Postilė by Bretkūnas:³⁵

(37)

Etymologically, gi derived from either the *g- or the *gh- particle of Proto-Indo-European, possibly from both particles which merged phonetically. The original meanings of *g- or*gh- remain relatively obscure, and Common Slavic *že, and both Greek de and ge are possible cognates (Dunkel 2014, 279–282). However, it is unclear how exactly gi came to function the way it does in contemporary Lithuanian, and more research is needed. It is clear that the particles of Lithuanian and German grammaticalized from different sources in different periods. More areal-typological as well as historical data is required to understand how different developments led to similar results in the two languages discussed in this paper.