Polish jakoby: an exotic similative-reportive doughnut? Tracing the pathway and conditions of its rise

1 Introduction

In modern Polish, jakoby is used as a particle or complementizer in contexts of hearsay. As a complementizer it is also frequently employed after verbal or nominal attachment sites (complement-taking predicates, CTPs) that denote the cognitive attitude (opinion, conviction, belief, doubt, etc.) of someone other than the speaker, including claims that are considered facts by some people; see (3).

Jakoby originated from a similative marker (jako ‘how’) univerbalized with the irrealis enclitic by. From the earliest attested stages (late 14th century) into the 17th century, jakoby was used as a comparison marker (similative) and as a subordinator of manner or purpose clauses. The former use has to some extent persisted, while the latter faded out (probably during the 19th century; see Section 4.2.2).

The development of jakoby presents an interesting case because it gives reason to assume that it diverges from earlier proposed pathways in various ways. At first sight, it is tempting to regard jakoby as an example that confirms the functional pathway of similative markers into evidentiality proposed by Gipper (2018); here and below > marks ‘sequence of appearance’, ⊃ ‘implies’:

On the basis of data from a limited number of languages, Gipper suggests that this pathway might be crosslinguistically common, but adds the caveat that this assumption must be tested on more languages (2018: 276–277). In fact, on closer inspection, the development of jakoby raises doubts whether Gipper’s pathway is the only one leading from similative to reportive marking. Actually, jakoby seems to have run through more than the pathway in (4) in acquiring a salient reportive function, but it seems as though jakoby “stripped off” functions in (4)’s middle part, while it retained the initial stage. Admittedly, loss of intermediary stages (with the initial one(s) retained) is not by itself a counterargument against some presumed pathway and simply yields a “doughnut category” (Dahl 2000: 10–11; Haspelmath 2003: 236). However, since there is practically no evidence for jakoby as an inferential marker, the acquisition of the now predominant reportive use should better be explained by other mechanisms. Moreover, jakoby’s diachronic changes are at variance with Saxena’s (1995) implicational hierarchy in (5) for various reasons (see Section 4). In fact, they are inverse to generalizations implied by this hierarchy, which was designed for say/thus (‘A < B’ means “A is lower on the hierarchy/appeared earlier than B”, thus “B implies A (but not vice versa)”):

Jakoby also lost its manner and purpose uses, and probably these uses are not part of the same developmental path (see Section 4.2.2.4). Furthermore, jakoby is only scarcely attested in contexts of quotation and after pretense-CTPs, although it has probably functioned as a subordinator since its earliest attestations, including uses as a potential complementizer, i.e., as a conjunction that identifies a clause as an argument of a predicative expression in an adjacent clause (cf. Kehayov and Boye 2016: 1; Wiemer 2021: 43–46).

This article outlines changes in the functional distribution and syntactic behavior of jakoby from Old Polish (14th–15th c.) to the present, in order to substantiate the objections and claims formulated in the preceding paragraphs. I will argue for a pathway that differs from Gipper’s in one crucial point, namely the role of inferential marking, and which, concomitantly, radically differs from Saxena’s. I will first comment on the available data and the state of research (Section 2) before I survey jakoby in modern Polish (Section 3) and then turn to its diachronic development (Section 4). The results will be discussed against the larger backdrop of recent research into the link between similative markers and evidentiality and the concomitant development of complementizers (Section 5). Examples are glossed to the extent necessary for understanding the morphosyntactic structure of the relevant clauses.

2 Data and state of research

The earliest attestations of Polish are from the late 14th century. The Old Polish period covers the 14th century up to about 1540, Middle Polish approximately covers the subsequent time until the late 18th century. Only very small corpora exist for Old Polish, but there are two dictionaries, Słownik Staropolski (SłStar) for the 14th–15th centuries, and Słownik Polszczyzny XVI wieku (SłXVI) for the 16th century, which contain extensive entries based on comparatively large bodies of texts. The reconstruction of jakoby’s way into reportive marking is based on the relevant entries in these dictionaries, on a critical assessment of existing research (Jędrzejowski 2020; Socka 2015; Wiemer 2015; Wiemer and Socka 2022), and on a combination of corpus samples (see Section 4). In addition, for the period inbetween, i.e., from the 17th century to the early 20th century, I investigated samples from the recently launched Electronic Corpus of 17th- and 18th-century Polish texts (up to 1772), abridged KorBa (<Korpus Barokowy), and from two smaller corpora covering the period 1753–1928; a sample of modern Polish was drawn from the PNC (see References). Of course, dictionary data (for the oldest periods, 14th–16th centuries) cannot be used for inferential statistics, and frequency counts have to be treated with caution when it comes to comparing them with corpus-based random samples (employed for the later periods). However, at present, this combination of analyses covering almost all periods of the attested history of Polish provides the best approximation for a reconstruction of the development of jakoby (as of other connectives) with the available resources.

3 Jakoby in contemporary Polish

This section illustrates the modern usage of jakoby as a particle or complementizer. It will be argued that jakoby has a primary reportive use and that evidence for an inferential reading is virtually lacking.

Jakoby is a word, i.e., it is a phonologically independent (non-clitic) lexical unit, used as a particle or complementizer with reportive or (more rarely) similative meaning. In both syntactic functions jakoby conveys epistemic distance, i.e., lack of full epistemic support (Boye 2012), which may range from epistemic agnosticism (see (1), (6)) to downward rejection of the truth of the proposition under jakoby’s scope (see (3)). This feature is induced by the incorporated irrealis enclitic -by, which explains the relation to irrealis functions, as well as why, as a complementizer, jakoby shows restrictions on the form of the finite verb and still serves as prosodic host for person-number clitics (see (2), (15) and Section 4.1). In both particle and complementizer use jakoby exclusively occurs in finite clauses, i.e., with verbs marked for tense and agreement (person/number). It scopes over propositions, that is, over claims (or questions) about situations (= states-of-affairs) that can be checked as being true or false, or which can be modified epistemically by the speaker (or another judging subject; see (10) and Footnotes 4 and 7), since the situation is assigned reference in space and time (Boye 2012: 192–198; Wiemer 2015: 227–228). Jakoby’s use as a similative marker has considerably receded and is being ousted by its cognate jakby. This observation suggests a shift from the domain of perception to the domain of cognition (knowledge and belief), which entails scope over propositional content.

Used as a particle, i.e., a function word with a flexible position that is not part of constituent structure, jakoby occurs with any finite form of the verb and almost never serves as a host of person-number clitics (see Section 4.1); see (6)–(8):

(8) furthermore demonstrates that semantic scope does not depend on the constituent to which jakoby adheres in linear sequence: though modifying the adjective niezwykłe ‘unusual’, jakoby implicitly scopes over a proposition that can be paraphrased as owe informacje były niezwykłe ‘those pieces of news were unusual’ (cf. Boye 2012: 183–184; Wiemer 2015: 228–229). As in the other cases adduced, jakoby indicates that this proposition derives from an earlier (somebody else’s) utterance.^[1]

Jakoby is thus sensitive to meaning components conveying propositional content of speech. As a complementizer, it predominantly occurs after verbal or nominal CTPs related to speech acts or cognitive attitudes (for figures see Section 4.2.2); see (9)–(13). It also occurs after verbs (or nouns) that do not, at face value, convey a speech-act-related meaning, but which reveal such a layer under specific circumstances. An example is negated wiedzieć ‘know’ in (9). It demonstrates that jakoby, as it were, “activates” a reportive meaning: the clause introduced by jakoby does not convey the headmaster’s lack of knowledge, but that the headmaster claimed to be ignorant of what is the propositional content of its clausal argument (the jakoby-clause). That is, jakoby marks reference to what the headmaster verbally claimed to be ignorant of, and this relates to verbal behavior.^[2] Thus, as a complementizer, jakoby cannot by itself introduce a speech-act related meaning, independently of the meaning of a CTP, but it can be used to make salient, or disambiguate, speech-act related meaning whenever the semantics of a CTP is compatible with such a meaning.^[3]

Importantly, reference to somebody’s cognitive attitude, or a “product” of mental activity (like wniosek ‘conclusion’ in (13)), is metonymically related to its verbalization: speech acts serve as external manifestations of people’s epistemic states (beliefs, convictions, doubts, etc.), however acquired or justified. By using jakoby, speakers refer to verbal utterances either directly or via cognitive attitudes that manifest themselves through respective utterances. These attitudes, or the related speech acts, are “labeled”, whereby the speaker qualifies them in terms of reliability or other properties relevant for epistemic support. This also pertains to alleged facts that people (other than the speaker) subscribe to: any opinion or alleged truth that the speaker refers to had to be uttered in some way or other to give it propositional content, and this content can be denied (see (15)).

Denial and skepticism against propositions that are either openly expressed or inferred by the speaker are frequent concomitants of jakoby; however, there may be more neutral stances. See, for example, (16), in which the speaker accepts the possibility that the messages (wieści) are true, and this is what raises his/her anxiety:

Thus, in case jakoby does not refer directly to (previously uttered) speech acts or their products (see (2), (3), (10), (11), (16)), it serves to ascribe cognitive attitudes (a.k.a. epistemic states) to other people (see (12)–(15)); in either case jakoby scopes over propositional content. That is, jakoby indicates that the speaker makes a claim about other people’s epistemic attitudes on the basis of their real or imagined utterances; I know of hardly any examples in which the speaker interprets just behavior, and in diachronic corpora, such examples are likewise exceptional (see Section 4). Crucially, jakoby does not point at any process or trigger of inference, as do inferential markers; instead, it scopes over the propositional “spell-out” either directly of hearsay or of attitudes or “products” of cognitive processes, which the speaker labels in such and such a way, and these labels are supplied by suitable CTPs. As a complementizer, jakoby connects a clause with the relevant propositional content to a higher-order predicate (CTP) which either directly relates to representative speech acts (i.e., assertions) or whose meaning potential includes a layer from which reference to representative speech acts can easily be activated. In this sense, jakoby and “its” CTP are in semantic concord: they share their evidential background of judgment, namely a verbal (communicative) one; but this concord concerns only the label, and the content of the judgment is specified (by the complement clause), while there is no connection with a possible trigger or justification of that judgment.

Consequently, even if jakoby does not refer directly to a speech act (“labeled” by the attachment site of jakoby), it is not used as an inferential marker in the sense employed in evidentiality research (Aikhenvald 2004; Speas 2018; Squartini 2008, among many others). According to the latter, inferential markers indicate that a proposition is based on inference, or more precisely: that the speaker makes a judgment about a situation (= SoA) S and that there is a more particular source of evidence which gives support for this judgment (i.e., that proposition p holds for S).^[4] S may be any situation and, as a rule, it does not refer to a speech event or assign a cognitive attitude to other people. The triggers (or backgrounds) of the inference vary, and usually the types of background provide the basis of classifications of inferential markers, which distinguish between perceptual triggers (or circumstantial knowledge) and encyclopedic knowledge (including knowledge about habits); to this we should add inferences triggered by hearsay.^[5] Diachronic pathways like Gipper’s are based on these assumptions so that an inferential stage is understood as a “mediator” between similative and reportive meaning.

However, Pol. jakoby does not target any such inference or its trigger (as either complementizer or particle): whatever the proposition (p) in jakoby’s scope tells us, it simply spells out other people’s assertions or, alternatively, mental states or products of intellectual activity that can be manifested by assertions. CTPs of jakoby label these assertions or related cognitive states, and jakoby “links” these labels to specific propositional content, but it does not point at, or specify, any process of inference or its triggers. In other words: the judgment modified by jakoby does not result from a speaker’s inference based on perceptual data or from their general knowledge background. Jakoby simply scopes over the content of speech or a related attitude of knowledge or belief, which the speaker assigns to another person. How this attitude was acquired, or why it was uttered, is irrelevant, but the speaker may distance themselves from the assertions of others and the stances behind them.

This makes jakoby differ also from what we normally observe in cases when inferences are based on hearsay, as in the following examples from English (17) and Spanish (18):

The conclusion (‘for a couple of days, IBM will be leaderless’) is based on a verbal announcement (‘John Akers says p’), but this conclusion is drawn by the author of (17), not by those referred to in this stretch of discourse.^[6] The author of (17) does not ascribe a cognitive attitude, or conclusion, to those referents mentioned in the discourse, but this is exactly what jakoby does (if it does not target speech acts directly). The case appears a bit more complicated in (18): on the basis of what Obama declared, the author of (18) concludes that Obama is ready to obey laws and civil rights and that this disposition will manifest itself in his conduct. Thus, one may ask whether the speaker’s assumption relates to Obama’s attitude or rather to his expectable behavior. Of course, these two interpretations are intimately connected (and may be variably fore- and backgrounded), either interpretation is compatible with parece que ‘(it) seems that’, and either interpretation amounts to inferential use. This differs for Pol. jakoby, which only ascribes attitudes to people, with the presupposition that they uttered something, or behaved in a particular way.

Table 1 summarizes these insights. Report-based inferential markers highlight the content of an inference (drawn by the speaker), with the triggering hearsay in the background, and the inference may relate to any kind of situation (not necessarily to another person’s cognitive attitude); see (17) and [1]. By contrast, jakoby either targets the propositional content of speech act(s) as such (thus behaves like a typical reportive marker; see [2b]), or it only assigns a cognitive attitude (or a conclusion, as the result of an intellectual process) to somebody else. In this case, there is a metonymic relation between speech act and cognitive attitude: the latter presupposes the former (attitude → speech act); see [2a]. Report-based inferentials are based on a metonymic relation as well, but this relation is better characterized as an implicature, and, above all, this implicature works in the opposite direction (speech act ⊃ inference about SoA (= p)), and the inferred p can relate to any SoA (knowledge background).

4 Tracing jakoby’s diachronic development

In Section 4.1, I start with an account of the origin and early history of jakoby based on extant research (mentioned in Section 2) and then present a corpus-based study covering the subsequent time from 1600 to our days almost without gaps (Section 4.2). In Section 4.3 I will provide the arguments in favor of an immediate diachronic link between similative and reportive use.

4.1 Characteristics of jakoby in Old and Middle Polish

As a word unit, jakoby consists of the connective (and interrogative) jako and the irrealis marker by. Since the earliest Old Polish attestations jako was a highly flexible unit, both in syntactic and in semantic terms. From a semantic point of view, comparison (similative) and manner were prominent at the earliest stages. The diachronic relation between comparison and manner function (manner > comparison or comparison > manner, or both from an indiscriminate meaning?) has remained unclear;^[7] however, in manner use jako is compatible with purpose, from which it cannot be clearly discerned in many discourse tokens of earlier Polish (see below). Apart from that, jako served as a complementizer (see the relevant entries in SłStar i SłXVI). In the latter function, jako died out, and in general, jako has been ousted by its cognate jak in practically all usage types,^[8] including the combination with by to mark comparison (see (19)–(21)). The details of these latter processes are an open issue and cannot be addressed here.

In turn, by originally was the 3sg-aorist form of byti ‘be’. In combination with the anteriority l-participle it marked a pluperfect. The aorist died out very early (only limited attestations in Old Polish), but the paradigmatically isolated form by survived as a 2P-enclitic (“Wackernagel clitic”); among its “favorite” hosts were clause-initial connectives like jako. However, =by could also attach to the l-participle, particularly after =by ceased to strictly abide by the 2P-rule. In any case, =by kept requiring the l-participle even after the latter turned into a general past marker (= l-form) and regardless of =by’s position and prosodic host.^[9] Alternatively =by could go with the infinitive, but attached to jako this option was restricted to manner and purpose (see below).

The combination =by + l-form (or infinitive) eventually turned into a discontinuous subjunctive marker;^[10] with the l-form, person-number enclitics always attached to =by yielding clusters like X=by=ś.2sg, X=by=śmy.1pl. Both requirements have persisted into contemporary Polish both in the “ordinary” subjunctive and for connectives with a final element -by, provided these function as subordinators (of adverbial or complement clauses); the full version of the PNC (1.8 bln segments) returns only 16 cases in which jakoby as a complementizer does not attract person-number clitics. The picture is diametrically opposed for jakoby as a particle (i.e., a function word with loose or no syntactic integration), for which queries in the full version of the PNC yield only 14 items with attached person-number clitics. Therefore, the remarkable point is that, in subordinator function, -by has remained this attractor, even though it can no longer be separated from jako and nothing can intervene (jako|by).

In sum, jakoby results from a morphologization process that has come close to morphological opacity, but the original collocation requirements of -by with the clausal predicate have remained and thereby turned into a semantically vacuous rule: the l-form only marks finiteness, but no tense; it can be characterized as a ‘morphone’, i.e., a property of word forms that bears no effect on syntax or semantic interpretation (Aronoff 1994). I gloss it lf.

Figure 1 summarizes this development. Steps (i)–(iii) present the rise of a new function word, whereas step (iv) indicates the split into complementizer versus particle. The latter process was accompanied by the retention of the l-form requirement and -by’s function as host for verbal agreement markers (→ complementizer) versus the loss of these properties (→ particle), and as a complementizer, jakoby always occurs clause-initially and typically right after its CTP. The chronological relation between steps (iii) and (iv) is another open issue that I will not tackle here.

Figure 1:

Jakoby’s morphosyntactic history in a nutshell.

Furthermore, jakoby (at different stages of its morphologization) through Old and Middle Polish was a highly flexible similative marker, as it could scope over constituents of practically any format, e.g., NP-internal modifiers (19) or participial adjuncts (20):^[11]

(19)

Okrvg yey oczy okolo zrzyenycze,
yakoby	[drogy-ego	yaczynct-a]
sim	precious-gen.sg	yakcynth-gen.sg
‘a circle around her eyes at the pupil, like [a precious hyacynth]’
(SłStar; Rozmyślanie o żywocie Pana Jezusa, ∼1500)

(20)

Sobota duchovna … rzeczona yest vtora pyrwa (…),
jakoby	[vczynyon-a	ze	wtor-ey	pyrwa]
sim	made-nom.sg.f	from	second-gen.sg.f	pirva
‘The spiritual Saturday (= sabbath?) is called the second first …, as if [it had been made first from second]’ (i.e., ‘… as if it had been made first Saturday from the second one’)
(SlStar; Rozmyślanie …, ∼1500)

Importantly, examples in which jakoby took scope over propositions, as in (21), are easy to spot, for instance when jakoby introduced a finite clause to compare some assumed state with another, hypothetical one.

(21)

Ach myloscz, czosz my vczinyla, eszesz me tak oslepila (…),
yakoby =ch	[nyk-ogo	na swecz-e	zna-l].
sim=1sg	nobody-gen	on earth-loc	know[ipfv]-lf-(sg.m)
‘Oh love, what have you done to me, you have blinded me (…) as if [I didn’t/wouldn’t know anybody (else) on earth].’
(SłStar; Piśmiennictwo polskie …, 1408)

Note that in similative use (and uses related to perception) jakoby can still be analyzed transparently as jako=by, with =by marking the subjunctive together with the l-form. Regardless of this, jakoby could take propositional scope if it syntactically modified units below clause level (see Section 3). Relevant examples can be found already in the earliest period; compare (22) with (8):

(22)

Kaza-l-y		zvyąza-cz	myl-ego	Iesus-a	povroz-my
order[pfv]-pst-pl		bind[pfv]-inf	dear-acc.sg.m	Jesus-acc.sg	tie-ins.pl
(…)	yakoby	[zlodzyey-a].
	sim	criminal[m]-acc.sg
‘They ordered to bind dear Jesus with ties,
(i) like [a criminal].
(ii) as if (he were) [a criminal].’
(SłStar; Rozmyślanie …, ∼1500)

Here zlodzyey-a agrees with the object NP myl-ego Iesus-a. This might be taken as a simple comparison of Jesus with a criminal (see (i)), but, possibly supported by the syntactic parallelism, the component -by favors a hypothetical statement (see (ii)), i.e., a suspended proposition: the proposition conveyed by the clause cannot be “checked”, because it lacks referential anchorage. Concomitantly, the original interpretation as a transparent combination of jako + irrealis clitic =by is weakened, since there is no verb (l-form) otherwise required by this clitic. For the same reason loss of transparency is inherent also with “objects” on subclausal levels (see (19)).

Other prominent uses were manner (23)–(24) and purpose, which sometimes interfered with a consecutive meaning (25). In manner function, jakoby introduced embedded WH-interrogatives (see 23).

(23)

Nye	vye-m,	y<a>ko.by =[chmy	y	mog-l-y	zatay-cz].
neg	know[ipfv]-prs.1sg	how.irr =1pl	3sg.m.acc	can-lf-pl	hide[pfv]-inf
‘I don’t know how [we may hide him].’
(SłStar; Rozmyślanie …, ∼1500)

(24)

A then Walthko, raczsza …, s Greglavem
scha	radzy-l-y,	yako.by	[gy	zaby-cz	mye-l-y].
rm	counsel[ipfv]-pst-pl	how.irr	3sg.m.acc	kill[pfv]-inf	aux-lf-pl
‘And this Waldko, the advisor …, counseled with Greglav (as to) how [they should kill him].’
(SłStar; Wiesz o zabiciu Andrzeja Tęczyńskiego, ∼1470)

(25)

Stro-g-my	skuthk-y	dobr-e,
build[ipfv]-imp-1pl	deed-acc.pl	good-acc.pl

gakobi =[chom	presz	ne	nasz-im	dusz-am
CONJ=1PL	through	3pl.acc	our-dat.pl	soul-dat.pl
otrzyma-l-y	sbauen-e].
receive[pfv]-lf-pl	redemption-acc
‘Let us do good deeds, so that, [through them, we (may) gain redemption for our souls].’ or: ‘… in order [for our souls to gain redemption].’
(SłStar; Kazania Gnieźnieńskie, 14th c.)

In manner and purpose uses, jakoby also occurred with infinitival predicates, as in (26):^[12]

(26)

Do	t-ego=m	záwʃze	by-ł	chętliw-y/
to	dem-gen=1sg	always	be-pst-(sg.m)	anxious-nom.sg.m

Iákoby	żywot	wieś-ć	vczćiw-y.
conj	life[m]-(acc-sg)	lead[ipfv]-inf	honorable-acc.sg.m
‘I always was anxious/to live an honorable life’ (‘… in order to live/as for how to live …’)
(SłXVI; Kochanowski, 1579)

As marker of manner or purpose, jakoby retained its transparent character: jako=[by + l-form/infinitive]. Simultaneously, these uses are associated with volition (intention), and for jakoby both faded out during the 19th century (see Table 5). By contrast, similative uses are more closely associated to cognition, and this favors an expansion into epistemic modality (which entails propositional content). Concomitantly, in similative use jakoby may retain its transparency even with propositional adjuncts (see (21)), but since it could always modify constituents of very different formats and, thus, be used without a verbal predicate, this flexible behavior supported loss of transparency (see (19)–(20)). In addition, it should be emphasized that orthography (joint vs disjoint spelling) cannot be regarded as a reliable indicator of (loss of) morphological transparency (jako=by > jako|by).^[13]

SłStar and SłXVI register some more minor functions, but no quotative uses. Moreover, before the 16th century, jakoby only rarely occurred clause-initially in the scope of predicates denoting speech acts and, if it did, these clauses mostly served to mark manner or purpose (cf. Wiemer 2015 for details). Simultaneously, in SłStar I did not find a single example with jakoby as a particle in the context of a speech act verb or any other indicator of reported speech. Similarly, Jędrzejowski (2020), who extracted 262 examples with jakoby from PolDi (see References list), found only two instances in which jakoby could be interpreted as a complementizer after verbs denoting speech acts (see further Section 4.2.2.1).

According to the SłXVI, the situation in the 16th century did not differ much, that is, during the 16th century reportive use of jakoby was still emergent, regardless of the syntactic function (complementizer or particle). Of course, one may question the representativeness of SłXVI (and of SłStar), and, anyway, even extensive dictionary entries do not create a valid ground for inferential statistics. Regardless, the huge amount of examples in SłXVI at least allows us to judge that, as a complementizer, jakoby still frequently occurred in similative function (27), probably not less frequently than as a complementizer after verbs (28) or nouns (29) denoting speech acts. SłXVI also adduces quite many embedded interrogatives with jako ‘how’ in transparent combination with =by (30):

(27)

ia milcżę/
á	cżyni-ę	iákoby =m	nie	widzia-ł.
and	act[ipfv]-prs.1sg	sim=1sg	neg	see[ipfv]-lf-(sg.m)
‘I am silent and act as if I haven’t seen anything.’
(SłXVI; Jan Leopolita: Biblia …, 1561)

(28)

oſkárżył Misboſetá ſyna Saulowego niewinnie/
iákoby	mia-ł	przeciw	krol-owi	mowi-ć.
rep.comp	aux-lf-(sg.m)	against	king-dat.sg	speak[ipfv]-inf
‘He accused Saul’s son Misboset unjustifiably that/as if he had spoken against the king.’
(SłXVI; Marcin Bielski: Kronika …, 1564)

(29)

Gdy	nowin-á,	iákoby	zeiś-ć
when	news[f]-nom.sg	rep.comp	depart[pfv]-inf
mia-ł	z	świát-á,	przyſz-ł-á (…).
aux-lf-(sg.m)	from	world-gen	arrive[pfv]-pst-sg.f
‘When the news that he had passed away arrived (…).’
(SłXVI; Piotr Ciekliński: Potroyny z Plauta, 1597)

(30)

Antipater	myſli-ł	iákoby	przez	iad
pn[m]-(nom.sg)	think[ipfv]-pst-(sg.m)	rep.comp	through	poison-(acc.sg)
mog-ł	zágładzi-ć	Alexandr-á.
can-lf-(sg.m)	exterminate[pfv]-inf	pn-acc
‘Antipater thought how he could exterminate Alexander with poison.’
(SłXVI; Historyja Aleksandra Wielkiego, 1550)

Notably, in SłXVI we find hardly any examples with inferential use. What we do find is jakoby clauses as clausal arguments of cognitive states, as in (31), which corresponds to modern usage (see Section 3).

(31)

wiele	ludz-i	proſt-ych	mniemái-ą/
many	people-gen	simple-gen.pl	assume[ipfv]-prs.3pl
iákoby	t-á	Mſz-á	wáſz-á
rep.comp	dem-nom.sg.f	mass[f]-nom.sg	your-nom.sg.f
by-ł-a	od	Krystuf-á.
be-lf-sg.f	from	Christ-gen
‘Many simple people assume that this your Mass was/is from Christ.’

Finally, SłStar and SłXVI give only scarce evidence for jakoby-clauses referring to (illusory) perception, for instance after zda(wa)ć się ‘seem’ or śnić ‘dream’; see however (27) and (32).^[14]

(32)

w ten cżas
mi.ſie	widzia-ł-o	iákoby	wrazi-ł
1sg.rm	see[ipfv]-pst-sg.n	sim.comp	press[pfv]-lf-(sg.m)
w mé ſerce miecż oſtry.
‘At this moment, it seemed to me as if he forced a sharp sword into my heart.’
(SłXVI; Baltazar Opec: Zywot pan Iezu Kriſta, 1522)

Jędrzejowski’s (2020) investigation does not give any corpus evidence for the 16th century (he sets 1543 as the beginning of the Middle Polish period, but the data from KorBa, which he used, start with 1600). Unfortunately, the 16th century remains a “dark period” for corpus-based studies in the history of Polish, after all. Consequently, claims concerning distributional patterns have to be treated with special caution not only for Old Polish (largely the time until 1500) with its very limited text sources, but no less for the 16th century.

4.3 On univerbation and the similative > reportive shift

Although presently without an ultimate proof, we may assume that cognition-based (i.e., evidential and epistemic) functions of jakoby developed on another path than the volition-based purpose (and the associated manner) function, and that, after the onset of the morphologization process of jako and by, this process bifurcated into two separate “branches” (see Figure 2). Moreover, there were conditions for a shift from similative to reportive usage that “bypassed” the inferential domain. These assumptions are plausible for the following reasons.

Figure 2:

Purported diachronic relation between volition- and cognition-oriented uses of jako=by > jako|by.

First, jako is one of the oldest interrogatives and clause connectives in Slavic with an enormously flexible (or vague) range of usages, among which manner (‘how’), quotes, but also the function of an epistemically neutral complementizer figure prominently.^[18] The addition of the enclitic =by (+l-form/infinitive) attenuated the reality status, but, as a marker of purpose or manner, this combination never ceased to yield a compositional reading: jako[=by + l-form/infinitive]. This differed for the similative use: although =by attenuated reality status, too, an advanced stage of morphological fusion (i.e., univerbation) ensued, as can be inferred already from those early and frequent instances in which jakoby modified constituents below the clause level, when no l-form of the verb was involved (see Section 4.1). In these cases -by was the only signal of irrealis functions. Despite the unreliability of joint/disjoint spelling (see fn. 14), a correlation between loss of morphosyntactic transparency and the shift from similative to reportive use is obvious, since jakoby has become an inseparable word unit and its reportive use became predominant; as a particle, even the l-form requirement is absent (see Section 3). That is, the compositional reading (jako = by) died out together with manner and purpose uses, whereas loss of transparency (jako|by) was possible very early in similative use. Presumably, this created favorable conditions for the non-transparent usage of jakoby to be “transferred” directly into its emergent reportive use, together with other factors (see below).

Second, as concerns the conceptual connection between comparison and perception with cognition, consider that comparisons can be drawn for objects of different ontological status (cf. Fortescue 2010 for discussion). In accordance with Lyons’ (1977) distinction between objects of first, second (= states-of-affairs), and third order (= propositions) – and adding illocutions as fourth order objects – similative markers can, in principle, be applied to objects on any of these four levels. If used to mark clause linkage, similatives may mark “propositional objects” either as adjuncts or as complements;^[19] see the examples in Table 8.

Table 8:

Types of ontological objects linked by similatives.

Similative (example)		Basis of comparison	Ontological status
(40)	My hands are like [emery paper].	First order	Physical object
(41)	Lying is like [betraying a friend].	Second order	SoA, event (can be observed)
(42)	He stumbled into the room as if [he were drunk]. → adjunct	Third order	Proposition (target of epistemic and evidential modifiers); related to cognition, but with suspended reality status
(43)	It seemed/He pretended as if [he were drunk]. → complement
(44)	She smiled at me like [“Hey, take it easy!”]	Fourth order	Illocution (target of quotatives), related to behavior

On this account, jakoby acquired the potential of scoping over propositions very early, and indeed we find such instances already in the oldest attested layers of Polish (see Section 4.1). There is no reason to assume that, in this development, purpose (or manner) uses of jako = by played any role as intermediate stages.

Now, what about connections between comparison, on the one hand, and reports and the assignment of epistemic attitudes, on the other? In most environments, irrealis markers, like -by, suspend propositions, in the sense that propositional content cannot be verified (nor falsified) since -by deprives it of specific referential anchorage. This suspension is tantamount to preventing, or canceling, a realis (a.k.a. factual) reading from the respective clause.^[20] Moreover, comparison (i.e., one of the original functions of jako) does not imply identity between two objects (of whatever ontological status, and expressed in whatever syntactic format). If comparison applies to propositional content, lack of identity (plus irrealis marking) easily leads to epistemic distance, i.e., lack of the speaker’s full commitment to the truth of p (this includes a possible agnostic stance). Thus, from a conceptual point of view, suspension of assertion supplies an immediate link between jakoby’s original similative function and its employment in propositional contexts.^[21] Before we finish this argument on how jakoby can have entered and become so entrenched in the reportive domain without the “aid” of an intermediate inferential stage, let us reconsider the more “traditional” reasoning.

The shift from perception to inference is well-known; as a reviewer remarked, this shift “consists in that the speaker observes a situation similar to one in which p would have occurred, and infers (by a process that has been called abduction) that p has probably occurred”. This is correct, but, crucially, inference, in particular abduction, is fraught with a varying degree of fault (“has probably occurred”), and this account of failure induces less than a full epistemic commitment to p being true. However, despite this conceptual link, the diachronic data discussed above does not support the assumption that this otherwise widespread shift has played any significant role in the history of Pol. jakoby. We also do not have any attestations of jakoby that would suggest a shift from report-based inferences to reportive meaning proper (for this possible link see the discussion in Section 3). Admittedly, there are occasional instances even from the late 19th century in which clausal complements headed by jakoby acquire an inferential function. One such example was supplied by a reviewer.

(45)

Odnosi się wrażenie, jakoby autor stosował swe uwagi do pewnych wzorców mniemanych władz duszy (…).

‘One gets the impression that the author applied his comments to certain patterns of assumed authorities of the soul.’ (1873)

Such examples cannot be denied, but they are isolated. Jędrzejowski’s (2020) corpus-based study as well as the much broader corpus study presented here show that such instances at all times (including the Old Polish data) have been extremely rare. There is no reason to assume that these findings are just incidental: why should any of the employed corpora (as well as the diachronic dictionaries) be considered to be less favorable of inferential than of reportive contexts? If, on a diachronic pathway, inferential meaning is assumed not only to precede reportive meaning but even to prepare the soil for its rise, we would expect inferential use to be more frequent than a reportive one at some point, in order to “ease” a transition from similative via inferential to reportive meaning. We may be less strict and admit that, under identical assumptions, salient inferential usage need not be frequent in the period immediately preceding the onset or increase in reportive use, but even then the available data does not meet the expectation. One may then try to argue (as did one reviewer) that a shift from similative (and perceptual) to inferential function already took place at the time of the earliest attestations of Polish (or even earlier), so that we no longer “see” these changes and the fading of its effects. However, apart from being quite untestable, such a hypothesis would drive us to two other problems: first, between the time when jakoby presumably shifted into inferentiality (before 1500) and the time when jakoby started becoming more salient in the reportive domain (17th century) would be separated by at least 200 years. Second, jakoby has kept its original similative meaning until now, albeit as a stylistically marked variant, but it was decidedly more salient through all periods in which there was but faint evidence of jakoby in inferential use.

All these facts considered do not make jakoby look like a unit with a “doughnut history”, i.e., with inferential meaning having “dropped out” after it created favorable conditions for reportive meaning to arise and thereby mediating a transition from its original similative function.^[22] We can, instead, assume that jakoby developed under specific bridging contexts that favored its speech-act related potential, but only rarely triggered inferential usage, and that previously emergent speech-act related uses of jakoby began to be amplified during the 17th century. This view on the facts would open up a reasonable alternative account of how jakoby became established in the reportive domain, while retaining its irrealis feature. As argued above, the comparison+irrealis meaning of jakoby can be “carried over” rather directly into propositional contexts, with an ensuing lack of full epistemic support. Apart from this conceptual link, the conditions relevant for emergent complementizer use since Old Polish times were favorable for reportive use, regardless of occasional inferential use. On the basis of these conceptual and empirical preconditions a scenario can be imagined that is compatible with the findings assembled in Sections 3 and 4. In particular, a switch into the reportive domain was possible because jakoby began to be employed as a means of taking a stance with other people’s opinions and assertions: the reporting speaker would keep their less than full epistemic support against other people’s judgments, whether expressed verbally or only implied. This usage – reportive plus epistemic distance – need not be inherited through intermediate perceptual-inferential steps as suggested in Gipper’s pathway (see (4)), or as assumed by Jędrzejowski (2020), in accordance with the mainstream of evidentiality literature.^[23] I, therefore, suggest that a shift toward reportive use was favored by the following context conditions:

1. jakoby more frequently occurred in the immediate vicinity of verbs, later also of other predicative units, that proved suitable as speech-act related CTPs, while such contexts with seem- and other perception verbs were much less frequent; if they occurred, occasional shifts into perception-based inferentiality proved insufficient for inferential meaning to become more entrenched.

2. Jointly, jakoby was characteristic in rhetorical contexts in which the speaker wanted to take a stance against another person’s assertion or viewpoint and to distance themselves from that opinion, for which the irrealis feature of jakoby proved suitable. The role of rhetoric requires a more thorough check; at present it is based not only on the discourse potential of jakoby in modern Polish (see fn. 2), but also on a repeated perusal of the diachronic material analyzed above.

3. An additional supportive factor was the metonymic link between cognitive stances and utterances that express them (see Section 3). No mediating inferential function was required, instead the similative source meaning could largely do without a perceptual background (= Gipper’s stage II) in order to apply to other people’s utterances and related epistemic attitudes.

This scenario takes the lack of full epistemic support as a constant feature of jakoby throughout its history (as does Jędrzejowski 2020 and as would comply with Gipper’s assumptions), but it avoids all problems in explaining why inferential usage obviously has not played any considerable role as a mediating stage in pathways discussed hitherto.

Figure 2 summarizes these considerations: jako|by indicates that the erstwhile enclitic =by turned into an inseparable segment of a new word (whereby univerbation was completed). We should remember that in similative (comparison) use jako(by) could often be read as a manner marker as well (see Section 4.1), and that the latter function is compatible with purpose (see (23)–(24)). The interrupted horizontal line is to indicate that the volition- and the cognition-based usage types belong to different developmental paths.

5 Conclusions

The facts analyzed for Pol. jakoby and the argument developed in Section 4.3 justify a diachronic pathway as presented in Table 9; it mainly accounts for the complementizer use. The underlying scenario stands as a working hypothesis.

Table 9:

Functional development of jakoby.

(i)	Comparison/similitude (jako), attenuated by irrealis marker (-by)	Objects are compared to types; comparison is flexible for objects of different ontological status (in Lyons’ sense)
		This flexibility makes possible a domain shift
(ii)	Lack of full epistemic supporta	(Commitment to) propositional content is suspended + speaker refers to other people’s assertions in order to take stance with them
		Metonymic shift
(iii)	Assignment of epistemic attitudes to other people (manifested by their utterances and/or behavior)	= [2a] in Table 1

1. ^aAs a side remark, the inner form of expressions like Eng. veri-similitude and its equivalents, e.g., in Polish (prawdo-podobieństwo) and Russian (pravdo-podobie) reflects the conceptual “mélange” of similarity and knowledge.

Stage (ii) spells out jakoby’s reportive function, while (iii) refers to its function in assigning epistemic attitudes with associated utterances. As pointed out in Section 3, it is crucial to realize that, if jakoby does not mark directly the propositional content of utterances (= hearsay), it never refers to inferences, but simply ascribes epistemic attitudes on the basis of utterances (or its behavioral equivalents); in other words: contemporary jakoby never refers to the evidential background of how a judgment was acquired, but only scopes over the propositional content of speech acts or metonymically related cognitive attitudes labeled by a predicate (verb or noun). Jointly with this, the analysis of diachronic data does not favor an assumption that behavior observed and interpreted by the speaker might have preceded utterances as an alternative basis for the assignment of attitudes.

I have furthermore shown that comparison markers need not turn into evidential (or epistemic) modifiers (e.g., particles or complementizers) in order to acquire propositional scope; rather inversely, the potential to scope over propositions seems to be a precondition for similatives to proceed on this path of functional development. For instance, cognates in other Slavic languages, like Cz. jako by, jakoby, Slk. ako by, akoby, only have similative and manner uses, in particular as complementizers after seem-verbs, and no evidential uses have developed. In fact, Pol. jakoby stands out for its reportive use at least on a European background (Wiemer 2018: 314–317, 321–328, 2022: 692–699). No comparable cases are mentioned in Boye et al. (2015), however some similative connectives in languages of Northeastern Europe do behave like reportive complementizers in certain contexts; cf. Holvoet (2016: 238–241) on Baltic and Kehayov (2016: 456–458) on Finnic.

Regardless, Pol. jakoby demonstrates that, in order for similatives to acquire a reportive function, there need not be an intermediate stage of inferential use. To understand this, we need to acknowledge that reference to other people’s epistemic attitudes (e.g., pogląd ‘opinion’), knowledge (e.g., prawda ‘truth’) and results of cognitive processes (e.g., wniosek ‘conclusion’), which a speaker assigns, crucially differs from reference to sources of judgment, which is what inferential markers specify (see Section 3).

Thus, Pol. jakoby is probably not a conceptual doughnut. It seems much more adequate to establish a direct link between similative and reportive function. This link preserves epistemic distance, or lack of full epistemic support (cf. Fortescue 2010), but it also “circumvents” perceptual and inferential uses. The irrealis component inherent to jakoby remained constant over all periods, but jakoby “survived” only in uses belonging to the cognition domain, while those related to volition/intention died out. This semantic development was probably favored by finite clauses.

Therefore, jakoby’s development as a reportive complementizer is not just a further “extension” of Gipper’s pathway from similative to inferential marking (as given in (4)); on closer inspection, the changes undergone by jakoby testify to a very different kind of development. Even more so, they contradict Saxena’s (1995) implicational hierarchy in (5) in many respects. The complementizer function was not the starting point, but based on jakoby’s initial function as a similative marker, which reverses Saxena’s hierarchy. Quotative function has been only scarcely attested over all periods and thus can hardly be ascribed any significance. Purpose clauses based on jako = by developed in different syntactic environments and clearly did not follow on complementizer use. In these two respects, parallels can be found between Pol. jakoby and a manner enclitic in Kambaata, a Highland East Cushitic language (Treis 2017), despite differences otherwise.^[24] Furthermore, no conditional function is known for jakoby. Finally, Saxena observed a movement from clause-level linker toward lower levels of constituency, but with jakoby the inverse is true: it has established itself as a propositional marker regardless of the syntactic format of the unit(s) which it modifies. Note, incidentally, that, while this is at variance with Saxena’s hierarchy, it conforms to Gipper’s analysis of similatives expanding into (inferential) evidentiality. Thus, her pathway turns out to be incompatible with Saxena’s generalization, too.

The reasons for these discrepancies may be sought not only, and not so much, in the different etymological sources of the expressions, but rather in the different syntactic environments in which these expressions changed, and in that they themselves differ in their morphosyntactic status. Pol. jakoby has become a self-standing word after an irrealis enclitic inseparably fused with a functionally vague clause connective in syntactically favorable positions. Among the cases surveyed by Gipper, Germ. scheinen and Du. lijken ‘seem’ have always been verbs (cf. also Mortelmans 2022); they have not turned into complementizers, probably because they have not depended on other verbs, but have themselves been capable of implying clausal arguments. The bulk of other cases discussed by Gipper represent clitics or affixes (with etymologies related to different modes of perception and/or conceptual backgrounds). So do, obviously, the majority of items examined by Treis (2017) from languages in Ethiopia (see also other contributions to Treis and Vanhove 2017). After all, the question stands whether Pol. jakoby, probably not being a doughnut, is an exotic species of reportive complementizer or whether analogous cases can be found elsewhere.