Quotative uses of Polish similative demonstratives

1 Introduction

Demonstratives are a broad and varied category but it may be said that, in general, they have a pointing function. They can either be accompanied by a literal act of gestural pointing, e.g., with one’s finger, gaze, or head, or fulfill a pointing function within the text by pointing to participants or propositions. When they specifically point to the following discourse, they are a powerful tool for demonstrating and establishing a joint focus of attention between the speaker and the hearer (Diessel 1999: Ch. 5). Example (1) illustrates this with the use of Polish taki ‘such/like this’ to introduce a discourse-new referent that emerges as the topic of the following discourse. These pointing and demonstrating properties make demonstratives effective quotative markers, as shown in (2), and indeed, demonstratives often grammaticalize to serve this function:

This article deals with Polish tak and taki, which are deictic demonstratives employed in a variety of uses. They are also ‘similative demonstratives’ (SDs) in that their “meanings lie at the intersection of the semantic dimensions of similarity and demonstration” (van der Auwera and Sahoo 2020: 1). SDs include, among others, German so, English such, Polish tak ‘so’ and taki ‘such’, and Spanish asi (cf. ‘demonstratives of similarity’ in König 2017; ‘similarity demonstratives’ in Umbach and Gust 2014). The combination of the similative and demonstrative semantics in the demonstratives in the above languages is to be contrasted with, for example, English like this, which realizes the two meaning components separately. In König’s (2015, 2017, 2020) terms, tak and taki have a deictic dimension (i.e., orientation, pointing) as well as the content (or ontological) dimension (i.e., manner and quality, respectively). Thus, with respect to these content distinctions, Polish adverbial tak may be labeled a manner SD (zrób to tak ‘do it like so/in this manner’), while adjectival (adnominal) tak-i may be labeled quality SD (taki samochód ‘such a car’).

As will be shown in this article, tak and taki are used in various quotative constructions, as in (2), and are instrumental in clause combining tasks involving the introduction of quoted discourse. As the subject is understudied for Polish, a more general aim of the article is to provide a corpus-based overview of strategies of direct quotation employing tak and taki. As will be emphasized throughout, the ways in which the SDs normally encode, respectively, manner and quality are also reflected in their quotative uses. Subsequently, special attention will be paid to verbless quotative frames of the type shown in (2), namely: (Conj) NP tak and (Conj) NP taki, to offer two grammaticalization-related accounts: one for tak and another for taki. I will argue that while (Conj) NP tak is a reduced clause (originally NP VERB tak), (Conj) NP taki is a stacking of two independent quotative strategies: (Conj) NP and taki. Although references will be made to equivalent quotative strings reported in other languages (e.g., German und er so ‘and he thus’; see Golato 2000), the analysis is largely focused on Polish.

The organization of the article is as follows. Section 2 introduces the concept of direct quotation. Section 3 overviews the system of Polish SDs in their primary functions. Section 4 addresses the relevant quotative strategies observed in the Spokes corpus and discusses methodological matters. The empirical comparison of the usage of tak versus taki, in particular their contrastive distributions, will provide important insight into the subsequent proposals as to how the two quotative markers have emerged. The status of (Conj) NP tak is discussed in Section 5. We will review arguments against a straightforward ‘say’-ellipsis account in coordinated structures in 5.1, and argue in Section 5.2 in favor of a reportative clause, complete with a predicate verb, undergoing a mix of grammaticalization, constructionalization, and pragmaticalization mechanisms. In Section 6, we will turn to (Conj) NP taki and propose that taki has grammaticalized as a lexical item, not as part of a clause, and may appear next to (Conj) NP, which is also another quotative construction in its own right. The article will be concluded in Section 7.

2 Introducing direct quotation

In this work, direct quotation (or simply quotation) refers to the representation of prior discourse from the perspective of the original speaker and the deictic center of the original speech situation, as in (3a), rather than from the point of view of the reporter, in (3b), which is a case of indirect reporting. The devices introducing a direct quotation, for example, John said/was like in (3a), will be referred to as quotative strategies.

Direct quotation and indirect speech reports differ in a number of ways. Some of the differences pertain to linguistic form, syntactic integration, illocution, and prosody. For example, in indirect reports, personal, temporal and spatial deixis is adjusted (I > he, your > Paul’s, is > was). Further, in indirect reporting the reported clause is embedded under the main clause predicate, thus yielding closer syntactic integration via subordination; many languages make use of that-type complementizers to mark this subordination relation. These deixis adjustments and syntactic embedding is absent in direct quotation. Note also that the quotative strategies in (3a) target a whole speech act complete with its illocution, while in indirect reporting, as in (3b), the original illocution is lost as speech acts are re-construed as propositions embedded under complement-taking predicates. Prosodically, to mark direct quotation, spoken language also typically resorts to intonational marking, e.g., pausing at the onset of the quote, as well as voice modulation or voice effects signaling a different participant entering the discourse (cf. Clark and Gerrig’s (1990) ‘delivery’ aspects).

Other differences between direct and indirect reporting are observed in the functional and pragmatic dimensions as well as in the nature of the quoted material. Non-linguistic or paralinguistic material may be quoted, such as gesture (She went, “<gesture>”), body stance, attitude, facial expression, voice quality or non-lexical sounds such as onomatopoeic sound effects, as in (4):

It is worth noting that quotation may also involve reports of words that were not spoken; instead, it may involve enactments of discourse that is only imagined and constructed (see e.g., Sams 2010). For example, one speaker’s indifferent silence might be creatively reported by another speaker in She was like, “I couldn’t care less” (cf. Tannen’s (1986) ‘constructed dialogue’ and ‘hypothetical discourse’). In other cases, reports of words actually spoken are constructed afresh. Such constructed discourse is marked by creativity, which in turn allows speakers to dramatically ‘perform’ the enactments and simultaneously add evaluative commentary (Buchstaller and Van Alphen 2012b; Clark and Gerrig 1990; Macaulay 1987; Wierzbicka 1974). Indeed, to be precise is not necessarily the objective in quoting, as shown by research into similative quotatives such as English like, whose comparative/similative semantics signals the approximative nature of the report (Buchstaller and Van Alphen 2012b: xv). This will also transpire in our discussion of Polish similatives.

Researchers have also pointed out more nuanced pragmatic effects achieved through the theatrical nature of direct quotation. They include: vividness (Chafe 1982; Wierzbicka 1974), a sense of immediacy (Tannen 1986), speaker involvement (Tannen 1982, 1986), speaker attitude (Spronck 2012), dramatization, personalization, intimatization, the marking of climactic points in the narrative (Golato 2000), reducing the speaker’s epistemic commitment to the reported event, interpersonal accommodation, convergence, and appeal for hearer involvement (Blyth et al. 1990; Buchstaller 2001; Ferrara and Bell 1995; Mathis and Yule 1994; Romaine and Lange 1991; Terraschke 2013).

3 Overview of Polish similative demonstratives (SDs)

Tak ‘so/in this way’ is an invariant item, while tak-i ‘such/like this’ is the corresponding adjectival SD, fully inflected for gender, case, and number (the citation form tak-i is the sg masc. nom. form). Tak and taki are closely related in their form and etymology: tak derives from Old Polish adverbial tako ‘so/in this way’, which in turn derives from Proto-Slavic pronominal *takъ ‘such’. This last form is also the source of present-day taki.

In Polish linguistic literature, tak is categorized as an adverbial pronoun (zaimek przysłowny) replacing adverbs and targeting manner or degree/intensity, and taki is categorized as an adjectival pronoun (zaimek przymiotny) replacing adjectives and targeting qualities (e.g., Fontański 1986: 12; Jodłowski 1973: 71–73; Miodunka 1974: 53–65; Padučeva 1992: 185, 188; Pisarkowa 1969: 61–69; Szczepankowska 2012: 282; Topolińska 2020: 57–59). Most sources more or less explicitly also consider both to be demonstratives (zaimki wskazujące), although one notable exception is Pisarkowa (1969), who classifies taki ^[1] as adjectival qualitative pronoun (zaimek przymiotny jakościowy), but not demonstrative (to be represented by ten ‘this’). Generally, the literature is limited. None of the works cited in this paragraph focus exclusively on tak/taki. Their scope covers some range, more or less extensive, of pronominal forms, and some merely categorize tak and taki without further discussion.

Pisarkowa (1969) and Miodunka (1974) are the most detailed descriptions of taki’s syntactic, semantic (and a few pragmatic) functions, although both accounts are still limited in depth as they deal with other pronouns as well. Additionally, Miodunka covers taki only in NPs, and tak is mostly left out in both works. Notably for this article, the quotative function is not reported.^[2] What Pisarkowa (1969) and Miodunka (1974) both discuss, among others, is the manner in which taki may be a marker of definiteness, on the one hand, or indefiniteness on the other (paralleled at least partly by definite and indefinite uses of English this). Miodunka also argues that the contrast of known/unknown to the speaker/listener is relevant, as illustrated in Example (1), where Mam takiego sąsiada ‘I have this neighbor’ introduces a referent that is specific (known to the speaker) but indefinite and discourse-new (i.e., previously unknown to the listener).

Besides the general consensus regarding tak and taki operating in, respectively, the manner and quality dimensions, similarity and similarity-based kinds are two other semantic components that are relevant here, and are referred to, for example, by Miodunka (1974: 54): takie auto (‘such a car/a car like this’) may be read as ‘a similar car (to this one)’ or ‘a car of this kind’, where the specific kind that the speaker has in mind may well be an ad hoc one (e.g., a car with blue stripes across the roof). By the same token, a speaker saying Ewa też tak tańczy ‘Ewa too dances like this’ while pointing at someone dancing means ‘Ewa dances like this/Ewa dances in a similar way’, once again indicating a similarity of sorts. Since tak and taki combine the dimensions of similarity and demonstration, they may be termed ‘similative demonstratives’ (equivalent to ‘similarity demonstratives’), as used by, among others, van der Auwera and Sahoo 2020.^[3] As has been argued for SDs in a number of languages, sets of objects related by SD-marked similarity are closely related to kinds in that “similarity classes created by manner and quality demonstratives are ad-hoc created kinds” (König and Umbach 2018: 297; see van der Auwera and Sahoo 2020 for a crosslinguistic study of SD-based ad hoc categories). Polish tak and taki fit this general crosslinguistic description fairly closely.

The two SDs also follow the crosslinguistically familiar pattern of demonstratives operating in two dimensions: the semantic content dimension (also referred to as ontological by König 2015, 2017; 2020; König and Umbach 2018) and the deictic dimension. The former is distinguished into manner, quality, and degree uses – all three being relevant to SDs specifically – as well as other categories such as place (here) or time (then), which are relevant to other types of demonstratives. Manner demonstratives refer to the manner with which actions take place (chodzić tak ‘walk like this’); quality demonstratives target qualities (taki samochód ‘such a car’); degree demonstratives pick up the degree of a property (tak wysoki jak ‘as tall as’). The deictic dimension is the specification of the position or distance of a referent relative to a center of orientation, typically the speaker and/or the here-and-now of the speech event. Languages vary in whether they distinguish between subtypes of the deictic domain, such as ‘proximal’, ‘medial’, and ‘distal’. Polish is one of those languages with a two-term distinction in identification demonstratives (ten – tamten ‘this – that’) and locative demonstratives (tu – tam ‘here – there’) but with no similar differentiation in the SDs tak and taki. Table 1 shows the formal differentiation of tak and taki in the ontological domain.

Other parameters that are relevant in demonstratives are associated with their deictic uses: the types and directionality of syntactic relations that demonstratives are involved in; as well as the discourse-pragmatic circumstances of the speech situation. The first distinction to be made is between exophoric and endophoric uses. While exophorically used demonstratives point to extra-linguistic entities, endophoric uses target referents or propositions in discourse, referring either backward to antecedents in prior discourse (anaphorically) or forward to referents in the following discourse (cataphorically).^[4] The uses isolated along this dimension are widely considered to be the hallmarks of grammaticalization processes, of which these uses are both sources and targets. König (2015, 2017, 2020) mentions further extended uses of demonstratives which are also ascribed to grammaticalization: boosters, quotatives, additives, causal markers, etc.

Below we will focus more closely on the interaction between exophoric, anaphoric, and cataphoric uses, on the one hand, and manner, quality, and degree semantics in Polish SDs. As is shown in Examples (5)–(7), tak and taki jointly cover the full range of combinations; namely, manner, quality, and degree uses are all possible exophorically, anaphorically, and cataphorically.^[5] We thus arrive at nine distinct configurations:

What emerges from (7a)–(7c) – and is indicated already in Table 1 – is that there is an overlap and interchangeability of tak and taki in degree uses across the board.^[6] This is in contrast to, for example, English so and German so, which cannot be replaced by their respective attributive quality SDs such and solch. Russian is also different in this respect as takoj, the quality and degree SD, cannot be replaced by the manner SD tak. Note that the interchangeability of tak and taki applies to adjectival or participial contexts such as (7), in which properties (and their degrees) are key elements; corresponding degree uses targeting adverbs normally select tak, as in (8):

However, even when the SD is followed by an adverb, there is still wriggle room allowing for interchangeability and subtle differences in emphasis, depending on whether the SD targets locally the adverb (tak Adv) and emphasizes the degree, as in (8), or whether it targets the Adv+Adj complex (taka Adv Adj) and emphasizes the property itself, as in (9):

Interestingly, the same flexibility in SD choice is to be seen in some extended uses deriving from degree readings of SDs. Let us consider booster and approximative uses. Boosters are degree SDs that emphatically denote relatively high scalar values. While tak is the only option for boosting actions or quantities, as in (10); for the modification of AdjPs, as shown in (11), tak and taki are interchangeable again, following the general pattern outlined in (7), in which an SD modifies an adjective by specifying a degree value of the property expressed in this adjective.

Similarly, approximatives are a type of SDs that denote an approximate scalar value usually expressed numerically (for this reason their use is somewhat restricted). With appropriate numeral-derived adnominal adjectives, once again, both tak and taki are fine, as in (12).^[7]

Importantly, the same formal conflation is not extended to hedging uses of SDs. As may be seen in (13),^[8] taki is the only option here, although one might expect tak to be able to modify an AdjP, as is the case in degree uses. We are thus faced with the following contrast: while degree, booster, and approximative uses allow for the free choice of tak or taki, hedging requires taki (when modifying AdjPs).

The explanation for the distributional contrast is the following. In hedging readings, the emphasis is on quality, not degree; in (13), the speaker is struggling to name the relevant property of the dog rather than ascertain the scalar value of the property. The SD to be used should be a narrowly focused quality-driven one.

Let us bring together in (14) the relevant distributional contrast pertaining to manner/quality/degree pointed out so far (anaphoric reference is not taken here into account since the discussion will ultimately lead to parallel observations about exophoric and cataphoric uses of SDs in quotation).

As may be seen, the formal differentiation and division of labor of the SDs is not clear-cut across the various semantic domains. Exophoric manner uses (14a) uncontroversially require the prototypical manner SD tak, and taki is ungrammatical. Exophoric quality uses (14b) require the prototypical quality SD taki. Degree uses and booster uses take both tak and taki (14c)–(14d). In turn, hedging uses involving properties (14e) can only be realized by taki, and the manner/degree tak is ungrammatical. Compared to degree, booster and approximative uses, hedging patterns with straightforward quality uses in that it is more geared to target the quality dimension (as speakers struggle to pinpoint a particular quality, not its scalar degree value).

These considerations of manner/quality semantics interacting with the extended uses of the SDs will also be relevant in the subsequent discussion of quotation by tak/taki.

4 Quoting with tak and taki: the corpus data

4.3 Results

4.3.1 General outline

Tables 2 –4 demonstrate the range of tak- and taki-strategies along with their relative frequencies. It is insightful to note that two verbless constructions – (Conj) NP tak and (Conj) NP taki – are the most frequent quotative strategies in the sample. Table 2 shows the basic types of quotation strategies (the left-side column) and the more specific subtypes including tak/taki that may be subsumed under the basic types. The subtypes are so-called stackings of more than one quotative device, e.g., the more basic strategy of (Conj) NP (i.e., conjunction + NP) is represented by (Conj) NP tak, (Conj) NP taki, etc. While Table 2 conflates tak and taki under the four larger categories, Table 3 and Table 4 treat the two SDs separately to better represent their respective contributions to the sample. Let us explain the shorthand notations used in the tables and illustrate the most important of the patterns observed.

Table 2:

Tak and taki conflated under basic types and subtypes.

Basic types	Specific subtypes (basic and stacked)		N
(Conj) NP	(Conj) NP taki	45	62
	(Conj) NP tak	13
	(Conj) NP taki że	2
	(Conj) NP tak coś tam	1
	(Conj) NP tak że	1
Report verb (‘say’, ‘ask’, ‘think’)	V tak	7	14
	V taki	4
	V że taki	1
	V jakoś tak	1
	coś tam V taki	1
Verb of associated action (AAV) (e.g., ‘look’, ‘approach’, ‘turn around’)	AAV taki	3	6
	AAV tak	1
	AAV że tak	1
	AAV coś taki	1
Other uses of taki	taki (modifier of quote)	6	7
	‘be’ taki	1
Total		89	89

Table 3:

Tak-strategies in the sample.

(Conj) NP tak	13
V tak	7
AAV tak	1
(Conj) NP tak że	1
V jakoś tak	1
AAV że tak	1
(Conj) NP tak coś tam	1
Total	25

Table 4:

Taki-strategies in the sample.

(Conj) NP taki	45
taki (modifier of quote)	6
V taki	4
AAV taki	3
(Conj) NP taki że	2
coś tam V taki	1
V że taki	1
AAV coś taki	1
‘be’ taki	1
Total	64

(Conj) NP (the self-standing nominal strategy) is the most frequent pattern in the sample (cf. Spanish y yo: “…” ‘and I: “…”’, Cameron 1998). It is widespread crosslinguistically (Buchstaller and van Alphen 2012a), also in its extended variety (Conj) NP thus, as illustrated in (15)–(16). The same quotative frame may be used mimetically to reproduce non-lexical material, as in (17)–(18).^[11]

(15)

German

Ich sagte ihm, dass er gehen muss. Und er so : “Ich werde es mir überlegen”.

‘I told him he had to go, and he (is/was) like, “I’ll think about it”’

(Kuteva et al. 2019: 402)

(16)

Portuguese

“Então o que quer dizer fornicar?” Pois ela assim : “Ai!”

‘“Well, what does fornicate mean?” And she (is/was) like: “Ai!”’

(Foolen 2008: 120)

(17)

Spanish

Y todo… todo el mundo de clase ahi (.) asi <a gesture of boredom>

And everyone in class (is/was) like <a gesture of boredom>

(Cameron 1998: 55)

(18)

German

Und er plötzlich so <gesture>

‘And suddenly he (is/was) like <gesture>

(König 2015: 42)

In Polish, for (Conj) NP, the nominal is typically a personal pronoun, and the construction is typically prefaced by the conjunctions i ‘and’, a ‘and-switch reference’ or to ‘so, then’. The function of the conjunction is usually to juxtapose the quoted discourse against a previous proposition or to juxtapose two speaker turns, hence the frequent occurrence of switch reference environments (X says “p”, and Y “q”), in which a conjunction is used to signal a topic shift in syntactically parallel clauses with different subjects. Consider Examples (19) and (20), with tak and taki, respectively (we defer a comparison of tak and taki till Section 3.3.2.).

(19)

a	on	taki	oburzony	no	nie	w	tym	momencie	na	mnie
and	he	taki	indignant	part	no	in	this	moment	on	me
i	tak	“to	ty	wolisz?	nie	to	ty	wolisz	słuchać	jakiejś
and	tak	so	you	prefer	no	so	you	prefer	listen	some
bajery	niż	prawdy?”
bullshit	than	truth
‘And he (is/was) kind of indignant, right? at me at that moment, and (is/was) like, “So you prefer…? no, so you prefer to listen to some bullshit than the truth?”’
(Spokes)

(20)

a	nagle	Maciek	do	mnie	“no	ale	ty	jesteś	najfajniejsza”	nie?
and	suddenly	Maciek	to	me	well	but	you	are	nicest-f	no
i	wiesz	ja	taka	“kuuurwa”
and	you.know	I	taki-f	fuck
‘And suddenly Maciek (says) to me, “Well, but you’re the nicest”, right? And I (am/was) like, “Fuuuck”’ [prolonged and high-pitched to indicate embarrassment]
(Spokes)

Another strategy in the sample is the typologically canonical technique of framing quotes with a finite report verb combined with an SD; verbs of speech, thought and cognition are all included here. Three such verbs are to be found in the sample: mówić ‘speak, say’, pytać ‘ask’ and myśleć ‘think’, all of them used in the historical present tense, which is instrumental in highlighting or foregrounding the reported discourse. The use of direct quotation in itself has also been reported to mark speech as foregrounded (with concomitant vividness) through dramatic re-enactment of events. Example (21) illustrates ‘say’+tak, while ‘say’+taki is shown in (22).

(21)

on	mówi	tak	“to	już	wychodzisz?”
he	says	tak	then	already	leave-2sg
‘He says like this, “Are you leaving already?”’
(Spokes)

(22)

no

i

w ogóle

stwierdziła

że(.)

no

jak

to

ona

zawsze

tak

well

and

all.in.all

concluded-3sg

that

well

as

part

she

always

so

wiesz (.)

yy

tak

yy

spontanicznie

bardzo

różne

rzeczy

robi

you.know

uh

so

uh

spontaneously

very

various

things

does

i

taka

mówi “a

w

sumie

to

tak

za

rok

to

and

taki-f says

and

in

sum

part

approx

in

year

part

mogliby

robić

już

sobie

dziecko”

nie?

could-3pl

make

already

refl

baby

no

‘And, all in all, she concluded that – well, as is typical of her, you know, she does various things very spontaneously – and she says like this, “And in fact, in around a year they could already try for a baby”, right?’

(Spokes)

Tak and taki may also be paired with what I call ‘associated action verbs’, i.e., verbs denoting an action that directly accompanies the speech event and describes a participant as performing it; in Spokes they include patrzeć ‘look’, podchodzić ‘approach’, odwrócić się ‘turn around’, and siedzieć ‘sit’, as in (23).

(23)

i	tam	gada	i	gada	i	gada	i	gada (.)	i	wiesz
and	part	talks	and	talks	and	talks	and	talks	and	you.know
ja	taka	siedzę	“no	świetnie”	nie?
I	taki-f	sit-1sg	well	great	no
‘And (s)he talks and talks and talks and talks, and you know, I (am/was) sitting like, “Well, great”, right?’
(Spokes)

Finally, another strategy type (the bottom of Table 2) is to use taki adnominally when the quote is treated as an integral sentence component, e.g., an argument of the predicate in (24). In cases like this, taki is a modifier of the quoted discourse rather than of the reported speaker, which is what differentiates this use from Conj NP taki.

(24)

a	wystarczyłoby	takie	“sorry	nie	przeszkadzam	ci?”
and	would.suffice	taki-n	sorry	not	disturb-1sg	you
‘And a simple “Sorry, am I disturbing you?” would be enough’
(Spokes)

There is general agreement that the semantic element of similarity/approximation present in SDs facilitates their quotative use – recall that quotations are typically non-verbatim, reproduced afresh (or constructed from scratch) and performed for the participants. This inherent nature of quotation is signaled by items that carry precisely these meanings, i.e., similarity, approximation and hedging. The signaling of this approximative nature of quotation is efficiently executed via SDs and would be comparatively difficult to encode in a report verb, which is likely part of the reason for the emergence of the verbless strategies.

SDs also share a demonstrating effect: they prepare the ground for the upcoming quotation and place it in the center of the hearer’s attention. The demonstrative nature of these words also goes hand in hand with the performative aspect of quotation, and they are especially useful for framing non-lexical quotations (gestures, facial expressions, sound effects, etc.), which, once again, would be difficult to introduce with report verbs (cf. Golato 2000: 31 for this observation in relation to German und er so).

Two verbless constructions – (Conj) NP tak and (Conj) NP taki – are the most frequently used SD-strategies in the sample. We will take a closer look at them in the remainder of this article.

4.3.2 Tak versus taki: insights from the corpus data

In many cases, tak and taki may be used interchangeably, although the two SDs do carry semantic undertones that go back to their primary quality and manner dimensions, respectively. This section outlines these nuanced contrasts, which occasionally surface in clear preferences as to which of the two SDs may be used or even in restrictions on their use. My focus will be on (a) taki characterizing the speaker, as in (25a), and (25b) tak modifying the action (implicitly) expressed in the quoted discourse, as in (25b):

(25)

a.

ja	taka “aaaaaahhhhh”
I	taki-f
‘I (am/was) like, “Aaaaaahhhhh”’
(Spokes)

b.

ja	tak “aaaaaahhhhh”
I	tak
‘I (am/was) like, “Aaaaaahhhhh”’
(modified)

In (25), the inherent quality/manner contrast is reflected in the sense that (25a) carries undertones of the quality dimension: taki characterizes the speaker, and the emphasis is removed from the speech event; the quote is thus likely to be a constructed representation of speaker attitude (=‘I was surprised/not in the mood to argue/etc.’). By contrast, (25b) carries undertones of the manner dimension: tak characterizes the action, i.e., the speech event, and the quote is thus likely to represent discourse that had been uttered previously.

This correlation between taki and constructed quotes characterizing the speaker on the one hand, and tak and manner readings introducing ‘real’ quotes on the other hand, is occasionally evident on closer inspection of the context. In (20), repeated below as (26), the context makes it almost certain that the quote is constructed from scratch and should receive a quality reading (=‘I was embarrassed/shocked’). On the other hand, the context in (19), repeated below as (27), indicates that the quoted words had been uttered, and tak-quotation should receive a manner reading:

(26)

a	nagle	Maciek	do	mnie	“no	ale	ty	jesteś	najfajniejsza”	nie?
and	suddenly	Maciek	to	me	well	but	you	are	nicest-f	no
i	wiesz	ja	taka	“kuuurwa”
and	you.know	I	taki-f	fuck
‘And suddenly Maciek (says) to me, “Well, but you’re the nicest”, right? And I (am/was) like, “Fuuuck”’ [prolonged and high-pitched to indicate embarrassment]
(Spokes)

(27)

a	on	taki	oburzony	no	nie	w	tym	momencie	na	mnie
and	he	taki	indignant	part	no	in	this	moment	on	me
i	tak	“to	ty	wolisz?	nie	to	ty	wolisz	słuchać	jakiejś
and	tak	so	you	prefer	no	so	you	prefer	listen	some
bajery	niż	prawdy?”
bullshit	than	truth
‘And he (is/was) kind of indignant, right? at me at that moment, and (is/was) like, “So you prefer…? no, so you prefer to listen to some bullshit than the truth?”’
(Spokes)

Admittedly, the interchangeability is still possible here, meaning that the observed pattern is only a tendency: in the sample, based on the context, in 22 out of 25 cases (88 %) tak-quotation seems to represent words actually uttered or gestures/noises actually made. On the other hand, taki-quotation seems to involve constructed representations of speaker attitude/state of mind in 69 % of cases. (It is difficult to quantify the trend with precision as some cases are difficult to categorize unequivocally.).

SD selection may occasionally be more restricted to an either-or choice. Let us consider Example (28):

(28)

coś	się	mu	wkręciło	i	taki	był	“ha ha hi hi”
something	refl	him	got.into	and	taki	was	<laughter>
i	nagle	tym	telefonem	tak	“jeb”
and	suddenly	this-inst	telephone-inst	tak	bam!
‘Something got into him, and he was like, “Ha ha hee hee”, and suddenly (he went) like, “Bam!” with his phone’
(Spokes)

Taki could not replace tak in i (on) nagle tym telefonem tak “jeb”; likewise, tak could not be used to replace taki in the first quote. The reason is again related to the quality/manner distinction: i (on) nagle tym telefonem tak “jeb” requires a manner SD to describe the manner with which the phone was thrown away, while taki był “ha ha hi hi” requires a quality SD to describe some properties of the speaker. While (Conj) NP tak may introduce any activity performed by the speaker, and this activity may be expressed as quoted constructed discourse, as in (28), (Conj) NP taki must introduce quoted (possibly constructed) discourse that directly characterizes the speaker. In other words, tak may target manners of performing actions, taki cannot. Conversely, taki is more suited for talking about the qualities of a participant, in line with its inherent adnominal character. Taki is thus geared towards quality readings, and tak towards manner readings, per their respective inherent properties.

The combination of the copula był ‘was’ and taki in (28) is also another indication that a quality reading is favored here ( ^ok był+quality); the use of the copula would not be possible in i on tak (*był+manner). The presence of the copula is also interesting for another reason: (28) may be seen as the kind of context in which hedging and quotative uses overlap. In both of these use types, the quality dimension is critical (see Section 2 for the quality dimension in hedging), and the combination of the copula and taki strengthens the inference that emphasis is placed on the qualities of the participant; thus i taki był “ha ha hi hi” is likely to be interpreted as ‘he was all happy, carefree, and in a good mood’. On the other hand, the modified tak-based version a on tak “ha ha hi hi” favors a more verbatim and action-based reading, one in which we can infer that the reported speaker did in fact laugh and that the action of laughing is what the reporting speaker is emphasizing.

The same distinction between manner-tak and quality-taki is clear in Example (29), where the negation context (nie odpowiedziałam ‘I didn’t reply’) makes it clear that we are dealing with either internal monologue or constructed representation of the speaker’s attitude or state of mind, not with actual speech.^[12]

(29)

nie	odpowiedziałam	tylko	się	odwróciłam	po prostu	już	taka
not	replied-1sg	just	refl	turned.around-1sg	simply	already	taki-f
“nie	mogę	nawet	na	niego	patrzeć”
not	can-1sg	even	on	him	look
‘I didn’t reply, I simply turned around, by then really like, “I can’t even look at him”’
(Spokes)

Note that the quotation in (29) is used adjectivally with a quality reading and may be replaced by other AdjPs, as in (30a). Tak, would only be possible with a degree reading (30b). Due to this distinctly adjectival-like status of the quote and quality-driven use of taki in (29), quotative tak would be an infelicitous replacement.

(30)

a.

Odwróciłam się po prostu już	quality/hedging taka zirytowana/wściekła.
‘I simply turned around, by then (kind of) irritated/angry.’

b.

Odwróciłam się po prostu już	degree tak zirytowana/wściekła.
‘I simply turned around, by then so irritated/angry.’

Another relevant observation is that taki, unlike tak, instead of referring to the reported speaker, may modify adnominally quoted discourse that is used in the sentence as one of the arguments of the verb, that is, syntactically the quote acts as an NP, as in (31)–(32). Note that taki is adjusted for gender to fit the neuter gender of the quote-turned-NP.

(31)

a	wystarczyłoby	takie	“sorry	nie	przeszkadzam	ci?”
and	would.suffice	taki-n	sorry	not	disturb-1sg	you
‘Something like “Sorry, am I disturbing you?” would be enough.’
(Spokes)

(32)

i	one	na	przykład	o	piątej	coś	przychodzą	rano
and	they	for	example	at	five	something	come-3pl	in.the.morning
nie?	więc	zawsze	takie	“dzyń!”	no	i	się	obudzę
no	so	always	taki-n	ding!	then	and	refl	wake.up-1sg
‘And they come for example at five-ish in the morning, right? So (there’s) always this “Ding!”, and then I wake up.’
(Spokes)

Compare the corresponding sentence below, based on (32), in which the quotation is replaced by an ordinary NP:

(33)

Więc	zawsze	jest	taki	odgłos	dzwonka.
so	always	is	taki	sound	doorbell-gen
‘So there’s always this doorbell sound.’

The quotations in (31)–(32) are syntactically equivalent to the NP in (33). In all cases, taki is used in its core quality meaning to modify an essentially nominal element. The adnominal position is obviously not available to the adverbial manner SD tak as it cannot modify NPs. The distinction between quality versus manner once again proves critical in the speaker’s choice between tak and taki. All in all, the contrastive use types of tak and taki – as illustrated in (28) and (31)–(32) – are schematized in (34). Admittedly, as already mentioned, in other examples the differences may be blurred considerably.

(34)

These underlying differences between tak and taki deriving from their, respectively, quality and manner uses will also be relevant in Sections 5.2 and 6, where we will deal with their respective grammaticalization paths.

5 What exactly is (Conj) NP tak?

6 What exactly is (Conj) NP taki? – the grammaticalization path of taki

In this section I review the grammaticalization path of taki as indicated by the combined factors of (a) the general properties of taki versus those of tak (see Section 4), (b) the empirical patterns of use of taki as shown in Table 4, and (c) the contrastive patterns of quotative use of tak and taki as discussed in Sections 5.3.2 and 5.2. Let us first review the most important conclusions made in Section 5.3.2.

Taki is inherently a quality SD and this meaning component is extended to its quotative uses, i.e., it emphasizes the properties of participants expressed in the quoted discourse. Consider the distinct readings produced by taki (49a) compared to tak (49b):

The above generalization is reflected in distributional differences: taki, as a quality SD, may be paired with the copula być, unlike tak, as in (25) reproduced below in (50). In such a case, być+taki in (50a) favors a quality reading. On the other hand, tak, unlike taki, may be used with the verb zrobić ‘do’. Such a combination, in (50b), strongly favors a manner reading. The concomitant use of być and zrobić with, respectively, taki and tak confirms and reinforces the usage preferences that are already present in the two SDs themselves.

Taki, not tak, is also the only choice in utterances where the quoted discourse is used as a nominal argument of the predicate verb and in which taki is used adnominally, as in (31), reproduced below in (51):

A close affinity of taki to hedging should also be noted. When taki combines with (Conj) NP, the construction is akin to hedging in that it emphasizes quality and may carry the hedging/approximative semantics. Crosslinguistically, quotative markers may overlap with hedging markers (Eng. like, Fr. genre) and have long been observed to introduce loose approximations or enactments of discourse (whether actually produced or constructed). This is also true of taki, which clearly has hedging uses also outside of quotation. From this it may be predicted that creatively quoting discourse that is constructed (e.g., attitudinal) will attract taki rather than tak as the quotative marker of choice because of its combination of quality and hedging/approximative semantics. This is the case in the Spokes sample, where taki seems to introduce constructed quotes in 69 % of cases,^[18] as in (52), where the constructed exclamation was not actually uttered, but is instead a representation of the speaker’s annoyance and exasperation. This was also the case in Examples (26) and (29), where the context shows clearly that the quoted discourse is a constructed representation of attitude.

With regard to the emergence of (Conj) NP taki, the starting point of the discussion is that in standard Polish the string has never been used quotatively in combination with ‘say’ or any other predicate verb: NP VERB tak/*taki: “Quote”. This is uncontroversial behavior crosslinguistically. Thus, quotative uses of taki may not be hypothesized as deriving historically from a clausal source construction schematized as NP VERB *taki; cf. again Example (31)/(51), which cannot be accounted for in terms of a clausal source construction. Instead, as an adnominal expression, taki belongs in the nominal domain (unlike tak, which – being an adverbial element – belongs in the clausal domain). Following this line of argument, I argue that taki has grammaticalized into a quotative as an adnominal item and independently of verb-centered clauses of any kind. It is an extension of taki’s quality semantics, which additionally has hedging/approximative uses, and this additional approximative component seems to be employed in quotative contexts to signal quality-based similarity. The findings that (a) the same items that are used as hedges/approximatives are also recruited as quotatives, and (b) direct quotes are typically approximations are well-known in the literature.

Once taki was available for direct quotation, it came to serve this purpose in a variety of environments, including its co-occurrence with (Conj) NP, which in itself is a widespread quotative strategy in Polish (Guz 2019) as well as crosslinguistically (Buchstaller and Van Alphen 2012a; Güldemann 2008). Hence the surface-form similarity of the two sequences: (Conj) NP tak and (Conj) NP taki. This is also the explanation for the four occurrences of ‘say’ + taki (including the reverse ordering taki + ‘say’) in Table 4: the grammaticalized quotative taki is very flexible as to what lexical and syntactic material it may co-occur with. In principle, however, the string ‘say’ + taki is non-canonical, colloquial and spoken, and should be distinguished as such from the canonical ‘say’ + tak.

There is more evidence for the different paths of development of (Conj) NP tak on the one hand, and of taki, on the other. As argued in Section 5.2, (Conj) NP tak has developed several types of reportative uses, including reporting discourse, movement, and states of affairs (Examples (42)–(45)). Except for reporting discourse, taki is not possible for these other uses, indicating once more a difference in origin. The distinct lines of development of tak and taki may thus be represented as follows:

(i)

7 Conclusions

Manner expressions are crosslinguistically recruited in clause-combining tasks involving the integration of direct quotes into speakers’ utterances. This is also the case in Polish. This article has shown that the way in which Polish tak and taki encode manner and quality is reflected in their quotative usage. In their ‘ordinary’ exophoric and endophoric uses, tak handles manner, taki quality, while both may serve to signal degree. In their extended grammaticalized uses, the manner/quality distinction carries over in that tak preserves the semantic ‘halo’ of manner, while taki preserves that of quality. In effect, there are fine-tuned differences in how a particular utterance may be interpreted depending on whether tak or taki is used (although one should allow for a degree of overlap and ambiguity). Thus a tak-prefaced quote favors a manner reading, i.e., focus is placed on the manner with which some activity was performed – e.g., a speech event or a physical action that is ‘quoted’, as in Example (28); this also invites an inference that the quoted discourse was, in fact, uttered (when speech is quoted) or that the action that it represents was performed (when physical activity is reported); a taki-prefaced quote favors a quality reading, i.e., the quoted discourse is to be taken as describing the participant (their attitude, mental state, etc., not necessarily their actual words), thus the quote tends to be constructed. This tendency is best illustrated in syntactic configurations requiring one SD over the other, with no wriggle room whatsoever, e.g., in the way tak and taki are compatible with, respectively, the verb ‘do’ and the copula ‘be’, as shown in (50), repeated below in (53):

Special attention was paid to the verbless strategies (Conj) NP tak and (Conj) NP taki. Despite their superficial similarity, two different underlying structures and two distinct development paths have been proposed, which also take into account the manner/quality distinction discussed above. It has been argued that (Conj) NP tak is a reduced reportative clause (originally NP VERB tak) that has emerged through a mix of grammaticalization, constructionalization and pragmaticalization mechanisms. On the other hand, (Conj) NP taki is a stacking of two independent quotative strategies: (Conj) NP and taki. The quality SD taki has been argued to have grammaticalized into a marker of hedging, approximation as well as quotation. As a quotative, it flexibly appears in a variety of environments, including its combination with (Conj) NP.