The diachrony of the Basque marker bait-: from a manner expression to subordinator

2.1 Overview

The morpheme bait- is usually described as a subordinator (see e.g., Artiagoitia et al. 2003: 711; Euskaltzaindia 1999: 12; Oyharçabal 1987). The main arguments in favor of the subordinate status of clauses with bait- are the following (Oyharçabal 1987: 263): (a) it is incompatible with other markers of subordination (relative -en or completive -la) or with the conditional prefix ba-, (b) the verbal forms with bait- cannot have allocutive forms (which are only possible in independent clauses). However, the degree of integration of the subordinate clause varies, as will be shown. Moreover, bait- is occasionally found in independent clauses.

Nowadays, it is used in eastern and northern dialects of Basque. However, old western and central texts contain a few examples of bait-, which suggests that the morpheme was once common to all varieties (Azkue 1923: Section 529; Lafon 1999 [1966]: 681; Lakarra 1986; Lakarra 1996: 191; Mitxelena 2011 [1981]: 530–531).

The marker behaves like a proclitic (Oyharçabal 1987: 257). No other morpheme can go between bait- and the verb. If a verbal particle is used, it has to be placed before it (Oyharçabal 1987: 259). The negative particle ez also precedes bait- and, since it ends in a voiceless sibilant, it causes devoicing of the initial consonant of the marker, which is sometimes reflected in writing, e.g., ezpaikara = ez ‘no’ + bait- + gara ‘we are’. It can attach to any independent finite verb: it cannot appear on imperatives, allocutive forms, verbs carrying the interrogative suffix or hypothetical verb forms (Oyharçabal 1987: 257). Several phonological changes occur between the final occlusive of the marker and the first consonant of the verb: -t + d-/g- = -t-/-k- (baitira, baikara), -t + z- = -tz- (baitzen), -t + n-/l-/h- = -n-/-l-/-h- (bainaiz, bailira, baihaiz). Additionally, Souletin has the form beit-, sometimes reduced to be- before n, h and l.

In subordinate clauses the marker is used as relativizer, complementizer and adverbializer. Its major uses can be divided into two classes (Figure 1). The classification I present here is roughly based on Oyharçabal (1987: 248).

Figure 1:

Functions of bait-.

First, bait- is employed together with several clause-initial elements:

1. conjunctions in adverbial clauses (reason (1), manner (2), concessive and result clauses are the main types)

2. interrogative pronouns in headed relatives (3), correlatives (4) and free relatives (5)

These kind of bait- clauses can be called ‘doubly-marked’ subordinate clauses, as there are two subordinating elements: bait- on the verb and a clause-initial conjunction or pronoun.

As regards the syntactic status of doubly-marked bait- clauses,^[5] except for free relatives, they are adjoined to the main clause, i.e., they are not embedded and cannot function as a constituent in the main clause. They occupy a marginal position with regard to the main clause (they can appear before or after it, depending on the clause type). Relative clauses can be extraposed (they do not have to appear immediately to the right of the head noun) and they can also appear as parentheticals in the middle of the main clause. In correlatives the degree of integration with the main clause might be considered higher (though they are not embedded), as the clause with bait- (which always precedes the main clause) is not easily omittable, and in the second clause there is usually a pronoun referring back to the clause with bait-. Finally, free relatives are embedded and function as a constituent in the main clause.

The prefix bait- is also used together with the conjunction ala in exclamatives (see Rebuschi [2008] for an analysis of their syntax):

In the second class (‘bait- only clauses’), there is no conjunction or any other element that determines the meaning. Thus, the interpretation depends on the context and often more than one reading is possible. Based on the function that the bait- clause fulfills, we can distinguish relative clauses (7)–(8),^[6] reason clauses (9)–(10), result clauses (which sometimes can be interpreted as expressing manner (11)) and complement clauses (discussed in Section 2.2). When the clause with bait- is placed before the main clause, the marker expresses reason, background information or topic (12).

Leaving aside complement clauses, the clauses with bait- illustrated above are adjoined and not embedded, similar to their doubly-marked equivalents. This also applies to examples interpreted as relative: the head noun and the relative clause do not form a single constituent and the bait- clause does not necessarily appear immediately after the noun it modifies. The only syntactic difference between relative and other bait- only clauses is that in the former there must be a participant shared between the main and the subordinate clause, while in the latter there is no such restriction.

Moreover, bait- can have a discourse function, expressing shades of consequence, result, contrast or emphasis (13)–(14). In the oldest texts, the verb marked with bait- can appear in the last clause of a sequence, and the clause with bait- can be linked with the preceding one using the conjunction eta ‘and’, as in (13).

Finally, the marker shows up in a few other areas of grammar: in place names, possibly in the animate locative postposition baita-, in the archaic prefix albait-, in indefinite pronouns and as a temporal conjunction.

Toponyms with bait- are found in some northern regions of the Basque Country, for example, Espela bayta ‘(the place) where the box tree is’ (1591) (Salaberri Zaratiegi 1996: 226). According to Salaberri Zaratiegi (1996: 226), they are Satznamen, which developed from bait- relative clauses. We can reconstruct (15) as the source construction with the subsequent ellipsis of the head noun.

The locative postposition baita- used with animate referents should also be mentioned here.

Mitxelena (2011 [1970]: 264) and Creissels and Mounole (2011: 179) mention the possibility that baita- could have developed from a subordinate clause (most probably a relative clause) with the verbal prefix bait-. According to Creissels and Mounole (2011) a change from ‘at the place where N is’ to ‘at N’s place’ could have taken place. From there it could develop into the locative postposition (see also Krajewska 2022), cf. ‘house/home’ > locative grammaticalization path (Kuteva et al. 2019: 233, 235–236).^[9]

Albait-, called prescriptif by Lafon (1980 [1944]), is an archaic verbal prefix no longer employed in Basque, found only in 16th- and 17th-century northeastern texts (Oyharçabal 1997: 61). As for al-, Lafon (1999 [1966]: 692) proposed that it is the particle of possibility ahal, as in ahal izan ‘can’. Albait- appears in second and third-person hypothetical verb forms and expresses an imperative that is to be executed in the future or if a condition is met (18) (Lafon 1980 [1944]: 491).

Furthermore, we find -bait as a suffix on a series of indefinite pronouns: for example, norbait ‘somebody’, zerbait ‘something’ or nonbait ‘somewhere’. Azkue (1923: 357) was first to realize that the subordinator and the suffix that attaches to interrogative pronouns are one and the same and that the particular variant used in subordinate clauses is also found in pronouns. Thus, in Soule the verbal marker became beit- (the first examples appear in the 17th century, but it became frequent in the 18th [Padilla-Moyano 2017: 703]), and the same form appears in pronouns too. In Biscay, the variant baist- is found as a prefix in a handful of examples in the anonymous collection of proverbs known as Refranes y sentencias (1596) and also occasionally in pronouns.

Finally, in some areas, especially in Soule, bait can be used as a temporal conjunction in non-finite clauses with the meaning ‘by the time’ (19). The affirmative adverb bai or forms built upon it – baiko or baikoz – are employed in some other dialects (Lafon 1999 [1973]: 44).

2.2 Complement clauses with bait-

The most common complementizer in Basque is -ela (Artiagoitia 2003; Artiagoitia and Elordieta 2016), which is used in indicative complements as well as in some subjunctive complements (-ela contrasts with -en, employed in indirect questions and some subjunctive complements). The range of use of bait- is more limited than that of -ela as only few verbs take complements with bait- apart from the more usual -ela complements.

The first class of verbs are factive predicates. As put by Kiparsky and Kiparsky (1970: 147), with factive predicates “[t]he speaker presupposes that the embedded clause expresses a true proposition, and makes some assertion about that proposition”. For instance, whichever sentence in (2) the speaker utters, he or she presupposes what is said in (21).

Predicates which can take bait- complements are mostly affective factive verbal and non-verbal predicates, e.g., pena/domaia da ‘it’s pity’ (22), harritzeko da ‘it’s surprising’, ez da dudarik ‘there is no doubt’, kontent izan ‘be happy’, xantza izan ‘be lucky’ (23). Epistemic factive verbs like ‘know’, ‘see’ or ‘realize’ never take bait- complements. Neither verbs of saying nor verbs of desire are attested with bait-. The marker also often surfaces with adverbs such as beharrik ‘luckily’, though examples without the subordinator on the verb are also possible (OEH,^[11] s.v. beharrik).

The predicates mentioned above are not uncommon in modern texts (22)–(23), but they are scarce in older sources. Before the 18th century, only a few examples are found, and they are usually ambiguous between reason and complement as in (24). This difference might be partially due to the type of sources available for earlier stages of the language (i.e., mostly religious texts), which probably did not favor the use of such predicates. Nevertheless, it seems that factive predicates tend to appear with non-finite complements in older texts (something still possible nowadays), as in (25).

Another class, quite different from factive predicates, consists of predicates such as gutitarik egin du or doi doiak egin du that expresses that something was very close to happening but eventually did not (26).

Furthermore, in some texts, the bait- clause appears to be the complement of a noun, e.g., of kausa ‘reason’ (27).

This appears to be a special case of a more general structure briefly described by Lafitte (1991 [1944]: Section 770), who suggests that any verb can take complement with bait- and the marker adds the sense of ‘the fact that’. In such cases, the subordinate clause is sometimes anticipated in the main clause by a demonstrative. In (28), for example, we have the demonstrative haur ‘this’. In (29), on the other hand, the main clause in the answer is elided. Such complements are also factive: as argued by Kiparsky and Kiparsky (1970), factive predicates can be paraphrased with the fact that (e.g., It’s surprising that … → The fact that … is surprising).

The final class are predicates of happening, such as gertatu or heldu ‘happen’. Contrary to emotive verbs, they appear already in the oldest texts (30). From the point of view of semantics, when such predicates are used the formally subordinate clause contains the foreground information (‘focal clause’ in terms of Dixon and Aikhenvald [2009]) and the main clause only emphasizes the statement. They can be classified as implicative verbs in the sense of Karttunen (1971a), that is, they imply the factivity of the subordinate clause. Factive predicates also have this property: the difference is that negated implicative verbs imply the negation of the complement (e.g., It happened that the poor one died → he died, but It did not happen that the poor one died → he did not die).

Thus, complement clauses with bait- tend to express facts or attitude towards them (Euskaltzaindia 1999: 1.3.6). This concords well with other uses of the marker. For example, reason and result clauses are factually oriented, i.e., they typically express something that has happened, as opposed to purpose clauses which express unrealized events (Hetterle 2015: 50–53; Thompson et al. 2007: 250–251). With regards to reason clauses, Pérez Saldanya (2020: 591) observes that bait- is used in clauses which are assertive or presupposed. Also, exclamatives can be seen as factive (see Grimshaw 1979), as they can be paraphrased with a factive predicate (“It’s amazing that …”, “I’m surprised that”).

Crosslinguistically, complement clauses tend to be embedded in the main clause and thus occupy a grammatical slot in it. In Basque, it is the case with complements with -ela. The situation is more complicated with bait- complements. Clearly, bait- complements are not easily omissible, which suggests tighter integration with the main clause (as compared to, for example, reason clauses). The order of clauses is fixed: -ela complements can precede or follow the main clause, but complements with bait- can only occupy the second position (Oyharçabal 1987: 249). In many cases, the clause with bait- is at the margin of the main clause: in (28) the link between the clauses is established with the demonstrative placed in the main clause which refers to the content of the subordinate clause. However, in examples such as (22) the subordinate clause might be analyzed as occupying the subject slot. Modern data (gathered in Norantz corpus [Oyharçabal et al. 2009]) add interesting details to the issue: some speakers put both clauses in single intonational phrase, but others place a break between the two clauses (31), suggesting that bait- complements are not always embedded even nowadays.

4.1 Etymology and functions of bai

According to Trask (1997: 209), “The word bai ‘yes’ is doubtless related to the affirmative morph ba-, as in badator ‘he/she’s coming’, formerly ‘he/she is coming’; bai may well be an ancient verb-form, along the lines of ‘it is so’, possibly involving the root -di- of the archaic verb *edin”. Lakarra (2018: 158) elaborated on this idea suggesting that the underlying verb form is *badadi, a form of the participle *edin ‘become’ and copular verb (subjunctive of ‘be’ in the modern language). It would mean something like ‘it is (so)’ or ‘let it be’. The prefix ba- could have functions similar to those it performs in the modern language: reinforcing finite verbs in clause initial positions and emphasizing affirmation. The affirmative function of bai might have appeared very early, but it is not the only function the particle has had. It seems possible that it came to be used as a kind of a manner expression, with the meaning of ‘so, as, thus, this way’.

This explanation has parallels in other languages: items with discourse functions frequently derive from verb forms. For instance, Bourdin (2008) analyzes grammaticalization of ‘go’ and ‘come’ into textual connectives and, in particular, devices expressing consequence of the previous discourse, or to the contrary, expressing something unexpected. Also several emphatic affirmative expressions similar to the reconstructed Basque phrase can be mentioned: Spanish (que) así sea ‘let it be so’, así es ‘it is so’ or French ainsi soit-il ‘let it be so’. Pérez Saldanya (2020) provides interesting parallels between bai, bait- and the prefix ba- and Latin and Occitan-Catalan forms. In Latin, the conditional si comes from the manner expression sīc (which in turn was reinforced as acsīc, which gave way to manner deictics in some modern Romance languages, e.g., Spanish así ‘this way’).

Moreover, other grammatical words in Basque can be explained through a grammaticalization of the verb *edin, e.g., baino ‘than’ or baina ‘but’ (Lakarra 2018: 109). Another particle, in a way more closely related to bai, is bada ‘so, thus’. The source of this particle could well be ba-da with the emphatic-affirmative ba and 3sg form of izan ‘be’. It would, thus, initially mean something like ‘it is so (indeed)’. If the source of bai was indeed a form or the verb *edin, such as *ba-dadi, bai and bada would have very similar sources (though probably different chronologies, bada being more recent).

König (2015: 41) examines manner deictic elements (such as so, this way, like) and various grammaticalization paths they typically move along and argues that renewal of manner deictic elements is very common. In many languages, the exophoric function is often lost: in English, for example, so, such or thus are rarely used exophorically, and this way or like this are employed instead. Partial renewal also took place, for instance, in Italian (ecco + si became così).

Anaphoric and cataphoric use develop from exophoric use: instead of pointing to the external world, deictic elements start to indicate parts of the discourse. As explained by König (2015: 43), manner deictics may develop into propositional anaphors and have whole clauses as antecedents, e.g., The meeting has been postponed? I suppose so. Anaphoric deictics can also turn into connectives expressing different circumstantial relations, e.g., causal (I didn’t like it, so I wrote to him.), concessive (John is very sick. Even so he goes to work) or conditional. Another well-documented change involving cataphoric demonstratives is the development of quotative markers (Güldemann 2008). Manner deictics are also frequently found in other grammatical markers, which are especially relevant for the discussion of Basque data: affirmative particles, additive markers, comparative markers and subordinators.

As regards affirmative particles, according to König (2015),

Affirmative particles typically derive from manner deictics. In the languages under discussion Italian si and French si (after negative questions) derive straightforwardly from the Latin manner deictic sic. English yes is the result of a fusion between yeah and swa and Polish tak means both ‘so’ and ‘yes’. Moreover, narratives can be confirmed by expressions like So ist es (German) or It must have been so and prayers, wishes and plans for the future by phrases like So be it or Ainsi soit-il (French) in a wide variety of languages (König 2015: 44).

The Basque bai has been used as an affirmative adverb ‘yes’ in answer to a question since the earliest texts. It can also reinforce affirmation (32), emphasize an opposition to a previous negation (33) or function as an adversative conjunction (34) (OEH s.v. bai).

The Basque bai can also be used in various coordinating-additive constructions. According to König (2015: 45), “(…) anaphorically used manner demonstratives develop into additive focus markers (Engl. also, Fr. aussi, Germ. ebenso, Swed. også, Russ. takže) and further into coordinating conjunctions”. Moreover, bai also signals ellipsis, usually of the verbal phrase. This is similar to the behavior of the English so (John is writing a book and so is MARY) or the French aussi (König 2015: 45).

While it is not very common, the particle bai on its own can be used as a conjunction meaning ‘as well as, and also’:

Another group of constructions involves eta ‘and’ and ere ‘also’ apart from bai. For example, bai is used in elliptical affirmative clauses together with the additive particle ere ‘also’. Bai is not obligatory, but there is a tendency to use it:

Importantly, bai cannot be used when there is no verbal ellipsis. This suggests that rather than expressing addition, bai signals that a part of the verb phrase was elided:

Interestingly, verbal ellipsis is also frequent in the emphatic use of bai discussed earlier (as in Examples (32), (33) and (34)).

Furthermore, there is bai eta, which means ‘and also’ (39). It is frequently contracted to baita, but it is attested in its full form (e.g., baieta ere in Etxepare). In modern Basque it is usually reinforced with the particle ere (OEH s.v. baita). However, there are numerous examples in texts without it.

Finally, in the oldest texts we find bai as standard marker in equative constructions (Lafon 1999 [1957]: 651–652). Equatives express equal extent and often consist of the following elements (Haspelmath and Buchholz 1998: 279):

In many languages equatives are similar to similative constructions, which express equal manner, e.g., He sings like a nightingale (Haspelmath and Buchholz 1998: 313). According to König (2015: 50), manner (or, more precisely, degree) demonstratives are used frequently as degree markers (German so gross wie Karl ‘as tall as Karl’), but sometimes also as standard markers (or both: as tall as Charles; as < alswa ‘quite so’).

In 16th-century Basque texts, bai is occasionally encountered as a standard marker in equative (41) and similative constructions (42). According to OEH, this use is not attested in any other text afterward.

To conclude the discussion of the functions of bai, we can say that it appears or appeared in several contexts in which manner deictic elements tend to be found according to König (2015): as affirmative particle, additive marker and standard marker. Though bai is not morphologically related to demonstratives (hau ‘this’, hori ‘that’, etc.), taking into account its etymology – *badadi ‘it is (so)’, ‘let it be (so)’ – it does not seem impossible that in the process of grammaticalization, it acquired anaphoric functions. This could happen because *badadi, uttered in response to a question, refers back to the content of that utterance, and even though there is no overt deictic element, it contains a zero anaphor.

4.2 From bai to bait-

In this section, I will put forward several hypotheses on how the marker bait- emerged and became a subordinator. The reconstruction proposed here for Basque can be divided in various stages, which are represented in Figure 6. In what follows, I will first focus on the development of the marker from the manner expression. Then I will try to explain how bait- started to be used in complex constructions and was eventually reanalyzed as a subordinator.

Figure 6:

Map of proposed extensions of bai and bait-. Subordinate clauses are marked with light gray color.