2.1 Variation

As pointed out above, exceptive clauses in Present-day Polish are headed by the morphologically complex C-head chyba że ‘unless’. I repeat the example given in (1) as (8) below:

The exceptive clause is finite. It consists of the finite perfective verb zdecyduje ‘will decide’ marked for the indicative mood. It is a preferred pattern in Present-day Polish. However and interestingly enough, the conditional clitic by can attach to chyba że:⁵

The presence of the conditional morphology on the complementizer triggers a morphological change. It turns the finite verb into an l-participle, a common Slavic participle form inflected for number and gender (for more details on complex tense forms in Slavic languages, see Migdalski 2006). Semantically, using by the speaker distances himself/herself from the exception introduced in the embedded clause, and expresses a commitment to a proposition that (s)he may feel safer to defend (cf. Krifka to appear). Although (9) is grammatical in Present-day Polish, it is not used as often as its counterpart with indicative morphology is.

Remarkably, exceptive chyba-żeby-clauses do not force by to occur adjacent to że ‘that’. It can also appear lower in the exceptive clause structure, i.e. either between the subject and the l-participle or on the l-participle itself, as, again, a slightly modified version of (1) shows (see Borsley & Rivero 1994 for more details on mobile inflection in Polish):

The different position of the conditional morpheme by does not give rise to two distinct interpretations, and the exceptive meaning remains unchanged. The next two corpus examples corroborate the claim that both patterns are still available in Present-day Polish:

Taken all together, we end up with three patterns marking an exception and containing the inherent exceptive complementizer chyba że ‘unless’:

All of them express a case in which an exception to a preceding (or following) statement will or may exist. As for (13a), I take chyba że to be a morphologically complex C-head occupying a single syntactic position. The second variant given in (13b) does not deviate from (13a). The only difference is that the conditional clitic by moves from a lower position and attaches to the C-head. Based on Migdalski (2006), I assume by to be a head base-generated in MoodP. Finally, when by occurs together with an l-participle, it is the l-participle that raises to by:

In the next two sections, I examine exceptive clauses in Polish adhering to the view that they should not be considered negative conditionals, and show that chyba-że-clauses have the status of peripheral adverbial clauses in the sense claimed by Frey (2012; 2016; to appear(a)). In doing so, I focus on (13a), i.e. the pattern with indicative morphology, as it is the most common pattern in Present-day Polish.

2.2 Exceptive clauses ≠ negated conditionals

The main objective of this section is to figure out how chyba-że-clauses differ from negated conditional clauses headed by the subordinator jeśli ‘if’. There are several reasons why negated jeśli-conditionals appear to be the most suitable adverbial clause type to be compared with exceptive clauses: i) they are adverbial clauses, ii) they exhibit mood alternation, and iii) – most importantly – they encode exceptions, i.e., the semantics of negated jeśli-conditionals is closely related to the semantics of exceptive chyba-że-clauses (cf. Clark-Clark 1977: 457; Brée 1985; Comrie 1986: 79; Declerck & Reed 2000; Leslie 2009; Nadathur & Lassiter 2014, inter alia). Concretely, what they have in common is negation of all alternatives being formed by substitution of the embedded clause by its alternatives, see Vostrikova (2018) for more details. While discussing selected differences of both clause types, I focus on, jeśli ‘if’, an inherent conditional complementizer disallowing any other interpretations (e.g. temporal) to avoid misunderstandings between any other adverbial clause types. Comparing them will therefore considerably help shed light on exceptive clauses. Based on the contrasts between both clause types, the question to what extent exceptive clauses in Polish are integrated into their host clause will be addressed.

Left periphery: One of the differences refers to the possibility of topicalization or focalization. Whereas chyba-że-clauses are strongly dispreferred on the left periphery of the matrix clause, negated jeśli-conditionals exhibit no restrictions in this respect:

A similar observation has been made by Dancygier (1985), who argues that Polish exceptive clauses – as opposed to their English counterparts – cannot be fronted:

If one is inclined to assume adverbial clauses to be derived by movement of an operator to the left periphery, as has been suggested in Geis (1970), Citko (2000), Demirdache & Uribe-Etxebarria (2004), Bhatt & Pancheva (2006), Haegeman (2012), to name but a few, one could argue that movement of an exceptive operator to the left periphery is blocked in Polish by an intervening element.

Non-assertive speech acts – questions: Another difference can be observed with regard to root questions. Only jeśli-conditionals can be in the scope of a question operator. Chyba-że-clauses are prohibited in environments associated with root questions:

This contrast illustrates that chyba-że-statements presuppose assertive force in the matrix clause, whereas their jeśli-nie-counterparts exhibit no illocutionary restrictions, leading to the conclusion that the exceptive clause cannot be part of a root question.⁶ Declerck & Reed (2000: 224) illustrate a similar contrast while discussing the semantics of nonirrealis unless-clauses in English (see also Fretheim 2006: 77 for another minimal pair):

Haegeman (2003: 322) shows – mainly based on conditional clauses in English – that they can be divided into two larger groups: i) event conditionals and ii) premise conditionals. Only the former, which are integrated into the host clause, however, can be in the scope of an interrogative operator. Accordingly, event conditionals are taken to be central adverbial clauses, while premise conditionals are treated as peripheral adverbial clauses. Frey (2012; 2016; to appear(a)) observes a similar contrast with respect to adverbial causal clauses in German, and claims that adverbial clauses that cannot become part of a question should be analyzed either as peripheral or as disintegrated adverbial clauses in the typology advocated by Haegeman (2003; 2010; 2012). It follows then that chyba-że-clauses cannot be central adverbial clauses, and at the same time the example (17a) illustrates that they cannot be considered disintegrated adverbial clauses either, as they clearly depend on the assertive force of the matrix clause.

The focus particle tylko ‘only’: Geis (1973: 245–7) observes for English that whereas if-not-clauses can combine with the focus particle only, unless-clauses cannot:

Polish patterns with English:

Geis (1973) argues that that the ungrammaticality of (19b) can be accounted for by postulating some constraint on lexical insertion of unless. This constraint seems to follow from the fact that whereas unless and chyba że express an exception to the validity of the matrix proposition, if … not and jeśli … nie, on the other hand, specify a condition for the matrix proposition. Now, the use of focus particles, i.e. only in English and tylko in Polish, presupposes the existence of a unique condition. This unique condition determines the possible world in which the matrix proposition is true. This is not the case in the exceptive clause where it is the matrix predication that determines the world in which the embedded proposition (= exception) is valid (cf. Brée 1985 and Declerck & Reed 2000). What is interesting about the focus particles is that they cannot occur within the matrix clause. In this context, the picture does not deviate from that one above:

These contrasts clearly illustrate that exceptive clauses – contrary to negated conditionals – cannot be in the scope of selected matrix clause operators. This, again, leads to the conclusion that exceptive clauses must attach at a higher structural position of the matrix clause.

Emotive predicates: It is well-known that conditional clauses can be used instead of declarative complement clauses occupying one of the argument slots of a matrix verb. Concretely, in (23a) the emotive matrix verb wnerwiać ‘annoy’ selects two arguments: The internal argument is the personal pronoun mnie ‘me’ marked for the Accusative case, the external argument, in turn, is a subject clause headed by the conditional complementizer jeśli ‘if’ (cf. Williams 1974; Fabricius-Hansen 1980; Pullum 1987; Quer 2002; Hinterwimmer 2010; Thompson 2012; Onea 2015; Schwabe 2016, among many others, for possible analyses):

Remarkably, exceptive clauses cannot be embedded under emotive predicates and occur as argument clauses, as (23b) convincingly shows. In this regards Polish patterns with English:

Factual conditionals: Factual conditionals (premise conditionals in Haegeman (2003)’s terms) contain a predetermined truth value. Compare the following example:

Uttering the B-sentence, the speaker presupposes that person A believes the content of the if-clause to be true. As shown by Declerck & Reed (2000), the use of negated conditionals in factual contexts does not express an exception:

We observe a similar contrast in Polish, too:

What negated factual conditionals express here is a result or a fact. Since this meaning is not compatible with the inherent exceptive complementizer, the ill-formedness of (27b) straightforwardly follows.

A final note is in order here about variable binding. The received wisdom has it that if a quantified NP can bind into a subordinate clause, it indicates that the dependent clause is integrated into its host clause. Accordingly, we expect chyba-że-clauses to disallow variable binding. However, (28) yields the opposite result:

This is surprising if we assume a quantifier to be able to bind an agreeing pronoun in the subordinate clause iff the quantifier c-commands the pronoun. Correspondingly, the exceptive clause in (28) is c-commanded by the quantified expression, i.e. by każdy pacjent ‘every patient’, which occupies the matrix Spec, TP position. Such cases clearly speak, at least prima facie, against the analysis proposed in the present article according to which chyba-że-clauses are taken to attach at a higher position in the matrix clause structure. However, examples like (28) do not pose a challenge because they instantiate cases of what has been referred to as modal subordination. Roberts (1987; 1989: 718) defines modal subordination as a “phenomenon wherein the interpretation of a clause α is taken to involve a modal operator whose force is relativized to some set β of contextually given propositions.” What this boils down to is that variable binding does not apply as a diagnostic test determining the status of chyba-że-clauses.

What we have seen so far is that in comparison to negated jeśli-conditionals, chyba-że-clauses appear ‘deficient’. Table 1 summarizes the main findings.

These differences strongly indicate that exceptive chyba-że-clauses cannot be treated as negated conditionals. The next section shows that chyba-że-clauses are peripheral adverbial clauses, and that they have the internal structure of root clauses equipped with ForceP.

2.3 Exceptive chyba-że-clauses as JudgeP modifiers

In this section, I discuss the external and internal syntax of chyba-że-clauses.

2.3.1 External syntax

As observed in the previous subsection, chyba-że-clauses are not sensitive to elements occurring in the matrix clause. This led us to the conclusion that they are not strongly integrated into the host clause. If this is the case, chyba-że-clauses are expected not to fall in the scope of a matrix negation marker. To illustrate two distinct scope relationships between a matrix negation and an embedded clause, compare the following minimal pair:

(29)	(Haiman & Thompson 1984: 517, ex. 13a–b)
	a.	They don’t beat us because they love us.
	b.	They don’t beat us, because they love us.

At the first glance, (29a) and (29b) do not differ syntactically. There is, however, one main semantic difference. Whereas in (29a) the embedded causal clause is negated indicating that beating takes place – not due to the fact that they love us, in (29b) no beating is involved and only the matrix clause is in the scope of the negation operator. Remarkably, chyba-że-clauses pattern with (29b), as a matrix negation cannot outscope a chyba-że-clause:

1. (30)

1. Nie
2. neg
1. przyjdę,
2. drop:by.1sg
1. chyba że
2. unless
1. mnie
2. me.acc
1. zaprosisz.
2. invite.2sg
1. ‘I won’t drop by unless you invite me.’
2. a) It is not the case [that I will drop by] unless you invite me.
3. b) #It is not the case [that I will drop by on condition that you invite me].

The narrow scope of nie in (30) provides evidence for the view that chyba-że-clauses are not sensitive to matrix operators and that they are not integrated clauses. In other words, they must attach above NegP. To determine the attachment height of chyba-że-clauses, it is reasonable to check if they are sensitive to higher modifiers.

Anand & Hacquard (2013) observe that embedding of epistemic modal verbs depends on the lexical semantics of a clause-embedding predicate. Concretely, they show that epistemic modal verbs can appear in complements of attitudes of acceptance, (31a)–(31c), but not in complements of desideratives or directives, (32a)–(32c):

(22)	(Anand & Hacquard 2013: 2–3, ex. 1a–2c)
	a.	John thinks that Paul has to be innocent.	(OKepistemic)
	b.	John said that Mary had to be the murder.
	c.	John discovered that Mary had to be the murderer.

(32)	a.	John wishes that Paul had to be innocent.	(*epistemic)
	b.	John wants Paul to have to be the murder.
	c.	John demanded that Paul have to be the murderer.

To account for this contrast, Anand & Hacquard (2013) divide attitudes, following Bolinger (1968), into two classes: i) representational and ii) non-representational, whereby only the former do quantify over an information state, e.g., a set of beliefs for believe, which epistemic modal verbs can be anaphoric to. Non-representational attitudes, in turn, are taken not to quantify over an information state. In this spirit, Anand & Hacquard (2013) propose Epistemic Licensing Generalization:

(33)	(Anand & Hacquard 2013: 3, ex. 4)
	Epistemic Licensing Generalization:
	Epistemic modals are licensed only in representational attitudes.

In this context, Lund & Charnavel (2020: 166) point out that concessive even though-clauses, for example, can be embedded under representational attitude predicates, but not under non-representational ones:

(34)	(Lund & Charnavel 2020: 166, ex. 21a–b)
	a.	John thinks that Paul went for a walk even though it’s raining (but it’s not actually raining).
	b.	John wishes that Paul would go for a walk even though it’s raining (*but it’s not actually raining).

The contrast between (34a) and (34b) shows that even though-clauses pattern with epistemic modal verbs. If they were embedded, the inference that the subordinate clause holds would be cancellable. This is clearly not the case. Czyba-że-clauses behave similarly. They can be embedded under representational attitude predicates (e.g. myśleć ‘think’), but not under non-representational ones (e.g. pragnąć ‘desire’):

1. (35)
1. a.

1. Jacek
2. Jacek
1. myśli,
2. think.3sg
1. że
2. that
1. dzisiaj
2. today
1. po
2. after
1. południu
2. midday.loc
1. mogą
2. can.3pl
1. wyjść
2. go:out.inf
1. na
2. on
1. wolność,
2. freedom.acc
1. chyba że
2. unless
1. sąd
2. court
1. zdecyduje
2. decide.3sg
1. inaczej
2. differently
1. (ale
2. but
1. tak
2. so
1. naprawdę
2. really
1. to
2. then
1. sąd
2. court
1. nie
2. neg
1. zdecyduje
2. decide.3.sg
1. inaczej).
2. differently
1. ‘Jacek thinks that they are allowed to be released from the prison today afternoon, unless the court will take a different decision (but in fact the court won’t take a different decision).’

1.  
1. b.

1. Jacek
2. Jacek
1. życzy
2. wish.3sg
1. sobie,
2. refl
1. żeby
2. that
1. dzisiaj
2. today
1. po
2. after
1. południu
2. midday.loc
1. wyszli
2. go:out.l-ptcp.vir
1. na
2. on
1. wolność,
2. freedom.acc
1. chyba że
2. unless
1. sąd
2. court
1. zdecyduje
2. decide.3sg
1. inaczej
2. differently
1. (*ale
2.    but
1. tak
2. so
1. naprawdę
2. really
1. to
2. then
1. sąd
2. court
1. nie
2. neg
1. zdecyduje
2. decide.3.sg
1. inaczej).
2. differently
1. ‘Jacek wishes that they would be allowed to be released from the prison today afternoon, unless the court will take a different decision (*but in fact the court won’t take a different decision).’

Lund & Charnavel (2020) assume that even though-clauses attach as EpisP modifiers, i.e., they adjoin to a higher functional projection, to Mod_epistemic in Cinque’s (1999) terms:

(36)	[frankly Moodspeech act [fortunately Moodevaluative [allegedly Moodevidential
	[probably Modepistemic [once T(Past) [then T(Future) [perhaps Moodirrealis
	[necessarily Modnecessity [possibly Modpossibility [usually Asphabitual
	[again Asprepetitive(I) [often Aspfreuentative(I) [intentionally Modvolitional
	[quickly Aspcelerative(I) [already T(Anterior) [no longer Aspterminative
	[still Aspcontinuative [always Aspperfect [just Aspretrospective [soon Aspproximative
	[briefly Aspdurative [characteristically Aspgeneric/progressive [almost Aspprospective
	[completely AspSgCompletive(I) [tutto AspPlCompletive [well Voice
	[fast/early Aspcelerative(II) [again Asprepetitive(II) [often Aspfrequentative(II)
	[completely AspSgCompletive(II) ]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]

Since chyba-że-clauses pattern with even though-clauses, I analyze them, in terms of Krifka’s (to appear) and Frey’s (2020) approach, as JudgeP adjuncts. They are not sensitive to negation or focus operators in the matrix clause, but they do depend on the assertive illocutionary force of the matrix clause, as they cannot modify non-assertive speech acts, and they are embeddable under representational attitude predicates. Following this line of reasoning, I argue that chyba-że-clauses attach at JudgeP of the matrix clause:

(37)

Now, I examine the internal syntax of chyba-że-clauses and argue that they project up to ForceP, as suggested in (37).

3.1 Etymology

Etymologically, the discourse particle chyba ‘presumably’ must have emerged out of the verb chybić ‘to miss’ / ‘to mistake’ / ‘to lack’ which is still present in PdP:

They can even co-occur which is mainly due to the fact that the verb chybić narrowed down its lexical meaning and cannot mean ‘to lack’ in PdP:

Furthermore, chyba could also be used as a preposition selecting noun phrases marked for the Genitive case:⁹

In (44) chyba has a narrow scope and selects the noun człowieka ‘man’. Its meaning can be paraphrased as ‘except’, ‘excluding’. Brückner (1927: 188) mentions the use of chyba as a noun ‘lack of’ which is supposed to have disappeared in the 16th century, and paved the way for the prepositional use. Remarkably, the use of chyba as a preposition disappeared from language use. In PdP poza selecting noun phrases marked for the Instrumental case is used instead of chyba:

In Section 4, I show that the use of chyba as a preposition was crucial for the development of exceptive clauses.

3.2 Old Polish (until 1543)

In Korpus tekstów staropolskich (‘Corpus of Old Polish Texts’) containing 17 texts put together by the Polish Academy of Science,¹⁰ I could not find any occurrences of chyba indicating that it is absent in the oldest sources, e.g. in Kazania świętokrzyskie ‘The Holy Cross Sermons’ or in Kazania gnieźnieńskie ‘The Sermons of Gniezno’. An independent search query in the PolDi corpus confirms this finding.¹¹ In total, the query yielded 51 occurrences of chyba in five sources, whereby only one of the sources can be considered Old Polish according to the periodization given in Table 2. Pamiętniki Janczara ‘Memoirs of a Janissary’ were written by Konstanty z Ostrowicy and are supposed to have appeared around 1500. The text contains three occurrences of chyba used as a discourse particle (= ‘presumably’). One of the examples is given in (46):

Chyba occurs in front of the clause and takes a wider scope, i.e. over the entire clause. It expresses speaker’s subjective attitude towards what is embedded.

3.3 Middle Polish (1543–1765)

To begin with, I discuss mp data from the PolDi corpus. In principle, in mp exceptive structures the conditional clitic by occurs adjacent to the complementizer że:

Exceptive structures also occur with infinitive verbal heads:

In few cases, though, the conditional morphology is missing. Instead, the indicative mood is used:

The examples presented in (47)–(49) clearly indicate that adverbial exceptive clauses existed already in the mp period. Apart from this, chyba can be also used in other contexts. Łukasz Górnicki (1527–1603) uses it not only in connection with że ‘that’ to introduce an exceptive clause, but also as a preposition:

In Listy do Marysieńki (1665–1683) (‘Letters to Marysieńka’) chyba is attested in 39 contexts. Table 3 shows its distribution. Chyba occurs as a discourse particle, a preposition and in combination with że ‘that’ as an adverbial complementizer.

The data extracted from PolDi resemble the use of chyba in another corpus.

In general, I extracted 46 cases from the KorBa corpus, also known as The Baroque Corpus of Polish.¹² An overview of how chyba was used in mp is given in Table 4.

Different patterns can be attested. But similar to the situation in PolDi, chyba is predominantly employed as a discourse particle:

Exceptive structures marked for the indicative mood outnumber their conditional counterparts:

Conditional exceptive clauses with the clitic by adjacent to the complementizer że occur, too:

Two additional remarks are in order here.

Firstly, what appears to be interesting is the fact that chyba can occur with other complementizers giving rise to an exceptive interpretation. In (54) the preposition chyba ‘except’ combines with the conditional / temporal complementizer gdy ‘if’ / ‘when’ and with the conditional clitic by. Taken together they render the meaning of the English complementizer unless:

Other complementizers can be attested, as well:

Instead of the complementizer że ‘that’, a wh-phrase can follow chyba:

Note that such constellations are not possible in PdP, neither with indicative morphology nor with conditional morphology:

This contrast reveals one considerable issue: In mp – as well as in op – the exceptive meaning does not come about the inherent complementizer chyba że ‘unless’, as is the case in contemporary Polish. Rather, it is due to the compositional meaning of the preposition chyba ‘except’ and the following complementizer that in older stages of Polish was not restricted to the declarative complementizer że ‘that.’

Secondly, chyba as a preposition can combine with the conditional clitic by and have both a narrow and a wider (= propositional) scope:

In both cases, chybaby should be analyzed as a preposition meaning ‘except (for)’. In (58), it combines with the PP u starych ‘by old (people)’. In (59), chybaby embeds a participial complement. Remarkably, in PdP chybaby can only have a propositional scope; it cannot scope over smaller constituents:

And even if it takes a propositional scope, it does not mean ‘unless’:

In other words, the example given in (1) cannot be used in connection with chybaby:

Based on these differences it is tempting to examine data from np to see how chyba-że-clauses developed.

3.4 New Polish (1765–1939)

I extracted and analyzed 177 chyba-cases from NewCor, a Corpus of 1830–1918 Polish.

The use of chyba as a discourse particle is strongly preferred:

In selected cases, chyba occurs together with a complementizer, giving rise to an exceptive interpretation. Both conditional and indicative morphology are attested:

As opposed to PdP, exceptive structures can still be introduced by chyba and conditional morphemes:

The cases (66)–(67) convincingly show that chyba could still be used as a preposition selecting CPs headed by a complementizer. I was not able to find any examples illustrating the occurrence of exceptives introduced by a wh-phrase.

3.5 Interim conclusion

We can recapitulate our diachronic findings as presented in Table 5.

What we can conclude from our diachronic findings is that exceptionality was expressed almost in all historical stages of Polish by means of the preposition chyba and the following complementizer which, in turn, could be a declarative complementizer, a conditional subjunction or a wh-phrase. I refer to these structures as exceptive structure in Table 5. Remarkably, they are not available nowadays. The prepositional use of chyba also disappeared from language use in the last century. The dedicated exceptive complementizer chyba że ‘unless’ must have emerged at the earliest in Middle Polish (1543–1765). Recall, in addition, that in Present-day Polish we have three patterns marking exceptionality on the clause level:

Diachronically, we can safely assume that (68b) occurred as first in the history of Polish. Then, indicative mood started to be preferred in the exceptive clause, (68a), whereas at the same time (68c) could be used, as well. As the next section shows, it is not surprising that the conditional clitic by contributed to the origin of the inherent exceptive complementizer chyba że ‘unless’.