Headless relative clauses in Early Irish

HEADLESS RELATIVE CLAUSES IN EARLY IRISH by Elisa ROMA I. Introduction: aim of the paper and overview This paper examines the distribution of headless relative clauses in Early Irish, drawing on a collection of relative clauses extracted from Corpus PalaeoHibernicum.1 The label “headless relative clauses” has been applied in the literature2 to different kinds of constructions: relative clauses without a nominal head in the matrix clause, including most prominently wh-relative clauses without a nominal or pronominal head, e.g. the English and Spanish examples in (1),3 called free headless relative clauses in Caponigro’s survey;4 relative clauses with light heads such as isolated determiners, which cannot otherwise occur as phrasal heads,5 e.g. the Spanish and the Old Irish examples in (2); and relative clauses without either a nominal head and a relative pronoun or co-referential element in the relative clause, as in the Maltese and Old Irish examples in (3) –termed superfree relative clauses by Caponigro6– where the clause itself denotes a referent.7 1) a. English I heard what you said. b. Spanish Tengo con quién hablar. have.prs.1sg with whom speak.inF ‘I have someone to talk to.’ c. English We will do whatever is needed. 1. StiFter et alii, 2021. 2. Dryer, 2013; Caponigro, 2021; Comrie and Kuteva, 2013b. 3. Examples are glossed according to the Leipzig Glossing Rules (https://www.eva.mpg.de/lingua/ pdf/Glossing-Rules.pdf). Words are separated according to modern editorial practices. Angle brackets are used for segmental pronominal morphemes that occur within the verbal lexeme and backslash is used in the glosses for initial consonantal mutations. Conventionally, conjunct particles, which affect verb stress and require conjunct inflection, have been detached through the symbol separating affixes, while conjunctions which do not trigger stress changes and are followed by verbs with absolute inflection have been detached by the symbol for clitics (=). Case letters are employed at the beginning of Old Irish examples and translations only when reporting the beginning of a gloss or sentence in the original text; the formulaic “.i.” which introduces many glosses has been generally omitted. 4. Caponigro, 2021. 5. Citko, 2004; Caponigro, 2021, p.19-22. 6. Caponigro, 2021, p.22-23; the Maltese example is from Sadler and Camilleri, 2018. 7. See Dryer, 2013, sect. 1. ETUDES_CELTIQUES_49_CC22_PC.indd 115ETUDES_CELTIQUES_49_CC22_PC.indd 115 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions 2) a. Spanish Lo que tú crees no es cierto. det.m.sg rel you.sg.sbj believe.prs.2sg neg is true.m.sg ‘What you believe is not true.’ b. indaas a ndubeir=som ar thuus than det.nom.sg.n n\rel\put.prs.3sg=3sg on beginning.dat ‘than what he puts first’ (Ml 47a14) 3) a. Maltese Għamil-t li għid-t-l-i. do.pFv-1sg comp say.pFv-2sg-dat-1sg ‘I did what you told me.’ b. ni-mbia durata ind iterum neg-obj.3sg.m\be.Fut.3sg rel\give.sbjv.prs.3sg in.3sg.acc.n[lat]again ‘he won’t have anything that he might give for it again’ (Ml 57a13) Early Irish does not have any wh-relative pronoun and therefore does not have free wh-relatives such as those in (1), except for free-choice whatever-relatives introduced by the interrogative pronoun cip.8 On the other hand, instances such as (3.b) occur in Old Irish, as already noted by Thurneysen,9 and will be the focus of this paper. Relative clauses with light heads as in (2.b) and other functional equivalents in Early Irish of headless wh-relative clauses, which will be termed here open relative clauses with set antecedents, will also be briefly tackled by way of contrast to headless relative clauses. Note that while Caponigro’s survey of headless relative clauses includes all the types listed above, the chapters on relative clauses in the World Atlas of Language Structures10 take into account constructions like (3) but not non-headed wh-relatives such as (1). The paper is organized as follows. Section 2 describes the corpus, the query and the process which led to the selection of headless relative clauses, and introduces the parameters which will then be detailed and used to classify examples in Section 3. Section 3.1 presents all the examples of headless relative clauses in the corpus and Section 3.2 analyses the data according to the syntactic and semantic parameters set out in Section 2, namely case roles in the relative and in the matrix clause, animacy and referential status. Section 4 briefly compares the distribution of headless relative clauses with the occurrences of similar relative clauses headed by a restricted set of antecedents, in order to complement the description in Section 3. Section 5 summarises the results and concludes the paper. 8. The greater cross-linguistic tendency for indefinites that express free-choice to be interrogative-based is noted in Haspelmath, 2013. 9. Thurneysen, 1946, p.315-316. 10. Dryer, 2013; Comrie and Kuteva, 2013b. 116 ELISA ROMA ETUDES_CELTIQUES_49_CC22_PC.indd 116ETUDES_CELTIQUES_49_CC22_PC.indd 116 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions II. Database and methodology The database used for this survey relies on Corpus PalaeoHibernicum (CorPH).11 CorPH is a freely available and searchable lexicographic database, which gathers at present 76 deeply annotated Old and Middle Irish texts. Out of a search in CorPH which extracted all textual units including a relative clause (search no. 4447),12 headless relative clauses and open relative clauses with set antecedents, i.e. relative clauses without any nominal head or with the pronominal heads listed in (4), have been extracted manually. 4) a (n-) ‘the’, cách ‘everyone’, cach óen ‘everyone’, cip hé ‘whoever’, intí (aní) ‘that one (thing)’, nech (ní) ‘any (one, thing)’, nach óen ‘anyone’, nanní ‘anything’, sechi(p) + stressed pronoun ‘who/whatever’ The search results of the original query were 5080 relative verb forms contained in textual units from 34 different texts. These include the Annals of Ulster, Baile Chuinn, the Poems of Blathmac, the Milan Glosses, the St. Gall Glosses and other glosses on Priscian, the Turin Glosses on Mark, the Armagh Glosses and many other minor glosses gathered in the Thesaurus Palaeohibernicus, and Monastery of Tallaght. For the full list of texts, their manuscript sources, editions and dating, I refer to the list of texts in CorPH and in the search results (see footnote 12): note that all Early Irish examples quoted in this paper rely on the online edition. Among the automatically extracted relative verbs, 537 open relative clauses were manually selected. Briefly, the selection criteria were: a) exclude all relative verb forms that do not match the definition of relative clause, that is “A construction is considered a relative clause… if it is a clause which, either alone or in combination with a noun, denotes something and if the thing denoted has a semantic role within the relative clause” (Dryer13); b)exclude all relative clauses with full NP heads; c) exclude relative clauses that occur right at the beginning of a gloss and can have a nominal antecedent in the Latin text. Criterion a)is meant to keep apart in particular complement clauses, which in Early Irish can assume the same morphological marking as relative clauses (see GOI,14 where Thurneysen notes that “In such constructions they are no longer relative clauses in the strict sense”). The 537 clauses that form the database of this survey belong to 15 different texts (Armagh Glosses, Baile Chuinn, Milan Glosses, St. Gall and Karlsruhe Priscian Glosses, Turin Glosses on Mark, Poems of Blathmac, Bern Servius, Karlsruhe Augustine, Karlsruhe Bede, Laon Cassiodorus, Munich Paul Commentary, Vatican Psalter 11. StiFter et alii, 2021. 12. The search results are retrievable, as all CorPH search results are, at https://chronhib. maynoothuniversity.ie/chronhibWebsite/tables?page=0&limit=10000&fprop=&fval=&dtable= search&ctable=&search=true&id=4447. Among the 78 texts listed in CorPH, the Epistle of Jesus and Cáin Domnaig have not been captured yet (as of June 2023). Note that both the automatic search and the manual selection may have overlooked examples, but the database is wide enough to grant representativity. 13. Dryer, 2013, sect. 1; see also Caponigro, 2021, p.51. 14. Thurneysen, 1946, p.318. HEADLESS RELATIVE CLAUSES IN EARLY IRISH 117 ETUDES_CELTIQUES_49_CC22_PC.indd 117ETUDES_CELTIQUES_49_CC22_PC.indd 117 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions Commentary, Vienna Bede, Monastery of Tallaght). Among these, 26 headless relative clauses of the type in (3.b) above –including a few doubtful cases– have been detected. Each of these instances will be discussed in detail in Section 3. The parameters used in Section 3.2 to describe their distribution are: • syntax: case-roles of the non-overt head Noun Phrase admitted in the matrix and in the relative clause, and their relationship (subject/nominative, object/accusative, prepositional, genitive) • semantics: animacy (animate/human vs. inanimate, cf. English who/what) and referential status of the Noun Phrase which includes the relative clause, with related inferences (definite-active, indefinite specific and indefinite non-specific, only partially corresponding to maximal, free-choice and existential free relative clauses respectively in Caponigro’s classification) • lexicon: pronominal heads, if any, admitted in relation to the syntactic and semantic parameters (relevant for open relative clauses with set antecedents) These parameters try to set apart the various features taken into account for the classification of headless relative clauses in the recent cross-linguistic survey by Caponigro15 and will be detailed below in Section 3.2. The analysis will not take into account the relativisation strategy (e.g. initial mutation, relative ending, relative particle with prepositions) since that depends on the verbs and on the case-roles, while its distribution does not show any relevant difference as opposed to fully headed relative clauses. III. Headless relative clauses in CorPH List of examples In this subsection I will present all the instances of headless relative clauses extracted from CorPH, listed in (5-27), and in subsection 3.2 I will comment on their features according to the syntactic and semantic parameters set down in Section 2. First the clearest examples are given (5-22), to which (3.b) also belongs, then instances for which other interpretations are possible (23-24) and doubtful cases (25-27) are presented. 5) as tormach hí tintud septien is tormach són cop.prs.3sg.rel addition.nom in version.dat Septuagint.gen cop.prs.3sg addition.nom that dano hí tintud teothis then in version Theodotio.gen ‘What is an addition in the translation of the Septuagint, that is also an addition in the translation of Theodotio’ (Ml 2a15) 15. Caponigro, 2021. 118 ELISA ROMA ETUDES_CELTIQUES_49_CC22_PC.indd 118ETUDES_CELTIQUES_49_CC22_PC.indd 118 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions As noted by Griffith,16 the emendation anastomarch in Thes.17 is unnecessary in (5). 6) amal as=n-ed as moam serc lin=nai like cop.prs.3sg.rel=rel-3sg.nom.n cop.prs.3sg.rel biggest love.nom for.1pl=1pl adchotadsam tri-ar saithar sain-diles síc is=ed rel\obtain.prF.1pl through-our work.acc proper-own.acc [lat]so cop.prs.3sg=3sg.nom.n as moam serc la dia maicc israhél cop.prs.3sg.rel biggest love.nom for God.acc son.nom.pl Israel.gen frissa-rusaithraigestar oc atuididen di cech imniud towards.rel-work.prF.3sg at their lead.vn.dat from every trouble.dat ‘As what we have most love for is what we have obtained through our own labour, so what God has most love for are the Children of Israel, for whom he has laboured leading them from every trouble’ (Ml 92c5) 7) is inunn dia-teit le=som an ioseph cop.prs.3sg same.nom to.rel-go.prs.3sg for.3sg=3sg det.nom.sg.nJoseph 7an effraim and det.nom.sg.nEphraim ‘that to which Joseph and Ephraim apply is the same according to him’ (Ml 100b9) 8) fo-theistib .i. forsa-moith cath són ɔdaig down-poured.dat.pl i.e. on.rel-break.prs.3sg battle.nom pron.nom.sg.nseek.prs.3sg loc ṅdia ditin place.acc acc for.his protection.dat ‘poured down, i.e. he who is defeated in battle seeks a place for his protection’ (Ml 110d10) Again, in (8) Thes. suggests to supply intí before forsamoith, but Griffith notes that this is unnecessary.18 9) int-í nad tabair digail is det.nom.sg.m-deict neg.rel give.prs.3sg punishment.acc cop.prs.3sg nach fitir side neg.rel.obj.3sg.nfind_out.pst.3sg 3sg.nom.m ‘He who does not inflict punishment, he is someone who does not know it’ (Ml 91a20) I adopt here for (9) the syntactic interpretation given by Griffith and taken up in CorPH:19 intí nad·tabair digail is a nominativus pendens phrase containing a relative clause, while nach·fitir, a headless relative clause, is the predicate of the main copular clause and side is the subject of the copular clause, resuming the nominativus pendens. This interpretation does not correspond to the translation in Thes.,20 namely ‘he who does not inflict punishment, it is that he knows it not’, but it is firmly based on the sense of the Latin text of which it represents a gloss –quomodo scit si 16. GriFFith and StiFter, 2013. 17. Stokes and Strachan, 1901, p.7. 18. GriFFith and StiFter, 2013; StiFter et alii, 2021. 19. GriFFith and StiFter, 2013; StiFter et alii, 2021. 20. Stokes and Strachan, 1901. HEADLESS RELATIVE CLAUSES IN EARLY IRISH 119 ETUDES_CELTIQUES_49_CC22_PC.indd 119ETUDES_CELTIQUES_49_CC22_PC.indd 119 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions non ulciscetur? Ut sequatur qui non punit nescit –and I therefore consider it quite uncontroversial. 10) ni-bía di mútaib bes huilliu in oen-sillaib neg-be.Fut.3sg of21 mute.dat.pl cop.sbjv.prs.3sg.rel more in one-syllable.dat ‘there shall not be any more of mutes in a single syllable’ (Sg 7a1) The relative clause in (10) is framed by the usual construction whereby comparatives appear in Early Irish as predicates; it has no nominal antecedent but the implied indefinite antecedent is modified by a partitive phrase (di mútaib). Partitive phrases with the preposition de/di can either follow or precede the relative clause, as the extraction of open relative clauses with set antecedents has clearly shown: in 12 instances the partitive phrase immediately follows the head and precedes the relative clause,22 in 7 instances it follows the relative clause.23 A possible head in (10) could be ní, which is frequently found in negative clauses and with comparatives (seeSection 4). 11) ni-mbia=adi ara-roibrea bith ma-ni eróima neg-obj.3sg.m\be.Fut.3sg=3sg.obj on.rel-use.pot.sbjv.prs.3sg life.dat if-neg receive.sbjv.prs.3sg óneuch from someone.dat ‘he will not have anything that he might live on if he does not receive it from someone’24 (Armagh Glosses, Thes. I 498, 15) 12) dare aliquis quod laborauit .i. as torad saithir do [lat]i.e. cop.prs.3sg.rel fruit.nom labour.gen to.3sg.m ‘to give what someone has worked on, i.e. what is the fruit of labour for him’ (Armagh Glosses, Thes. I 498, 16) 21. According to Bauer etal., 2017, and CorPH, di here is the feminine form of the numeral ‘two’, and mútaib a dative dual. However, if this were a dative phrase governed by the comparative, the word order would be very awkward, so I conclude that the meaning ‘than two’ must be inferred from the Latin text the gloss is about, i.e. nulla enim syllaba plus duabus mutis se iuxta possitis nec plus tribus consonantibus continuare potest. The interpretation maintained here is also implied by the translation in Thes., which puts ‘than two’ within brackets. Whatever the analysis of di mútaib, anyhow, bes huilliu is a headless relative clause. 22. One example from the Poems of Blathmac, 6 examples from the Milan Glosses, 5 examples from Monastery of Tallaght: text unit IDs in Corph S0005-91, S0006-2709, S0006-3520, S0006-447, S0006-5961, S0006-5961, S0006-7449, S0070-149, S0070-153, S0070-288, S0070-640, S0070-88 (when CorPH does not give the locus for a text unit, I give the text unit ID for its retrieval). 23. 3 examples from the Milan Glosses and 4 from Monastery of Tallaght: text unit IDs in CorPH S0006-5756, S0006-7376, S0006-7469, S0070-341, S0070-40, S0070-40, S0070-411. 24. Thes. translates ‘he will not have it to enjoy unless he receive it from some one’. However, the context is noch is fírfoirbthe qui inplet quod dictum est uade et uende omnia, and CorPH rightly translates ‘That is to say, He is truly perfect who fulfills that which was said, ‘Go and sell all’; He will not have anything to enjoy unless he receive it from someone.’ 120 ELISA ROMA ETUDES_CELTIQUES_49_CC22_PC.indd 120ETUDES_CELTIQUES_49_CC22_PC.indd 120 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions 13) adchos[a]ire .i. adchois maldacht in popoil foir averter.nom i.e. rel\avert.prs.3sg curse.acc det.gen.sg.mpeople.gen on.3sg.m ‘an averter, i.e. one who averts on himself the curse of the people’ (Turin Glosses, Thes. I 492, 111) 14) 7ni cian huaib imbíat and neg long.nom.nfrom.2pl in rel\be.prs.hab.3pl ‘and it is not far from you where they are usually’ (Monastery of Tallaght)25 The following instances (15-19) group cases of imprecations, curses and beatitudes, where the headless relative clause depends on mairc ‘woe, alas’ or céin·mair ‘happy’; it therefore occurs where a head noun in the accusative (or in the dative preceded by do) vs. in the nominative, accusative or dative would be expected respectively. The construction with a relative clause is explicitly noted in eDIL26 for mairc (s.v. mairg, mairc, “c. with omission of subst., its place being taken by a rel. sent.”) and also figures in the first example reported in eDIL for céin·mair, viz. céinmair noda-ainsed ‘happy he who should spare him’ (Togail Bruidne Da Derga 111, eDIL s.v. maraid, I b). All instances in CorPH happen to be in poetry; (16) and (17) contain two relative clauses each. 15) Mairg adchondairc mac Dé bí woe rel\see.prF.3sg son.acc God.gen alive.gen.m fri croich in-a glóeth-rigi against cross.acc in-his glue-stretch.vn.dat ‘Woe to anyone who has seen the son of the living God stretched fast to the cross.’ (Poems of Blathmac 481-82) 16) Mairc rochar mac ríg nime woe rel\love.prF.3sg son.acc king.gen sky.gen adchondairc achró-lige rel\see.prF.3sg his m\blood-lie.vn.acc ‘Alas for the one who has come to love the son of the King of Heaven, who has seen him lying in gore’ (Poems of Blathmac 529-30) 17) Céin-mair ro<d>creiti, ro<d>car long-live prF<obj.3sg.m.rel>believe.prF.3sg prF<obj.3sg.m.rel>love.prF.3sg in coimdiu rocoínsemmar det.nom.sg.mlord.nom lament.prF.1pl.rel ‘Happy who has believed him, who has come to love him, the Lord we have keened’ (Poems of Blathmac 533-34) 25. Text unit ID in CorPH S0070-398. 26. eDIL 2019. HEADLESS RELATIVE CLAUSES IN EARLY IRISH 121 ETUDES_CELTIQUES_49_CC22_PC.indd 121ETUDES_CELTIQUES_49_CC22_PC.indd 121 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions 18) Céin-mair rochúalae clais cóir long-live rel\believe.prF.3sg assembly.acc choir.gen síl n-Ádaim im-a senóir seed.acc acc-Adam.gen around-its ancestor.acc ‘Happy who has heard the choir assembly, Adam’s seed around their ancestor’ (Poems of Blathmac 725-26) 19) Scarfaid – mairg nád-ais athnú– separate.Fut.3sg woe neg.rel-fear.prs.3sg his m\envy.acc caírcha gela fri mindu sheep.pl.acc white.pl.acc against lamb.pl.acc ‘He will separate –woe to who does not fear him– white sheep from lambs.’ (Poems of Blathmac 1067-68) A case which is similar to (15-19) is (20), where the main clause also apparently consists of a nominal predicate. Although CorPH gives the translation I report below and analyses bebais as a relative verbal form, other interpretations are possible; for instance, Murphy27 includes in the sentence the previous words bess tress and translates ‘perhaps in the third month over a year he shall die by the sea.’ 20) mi for bliadni bebais muir month.nom on year.dat die.Fut.3sg.rel sea.dat ‘a month and a year (for the one) who will die in the sea’ (Baile Chuinn)28 In (21), (22) and (23) the same copular predicate sain ‘else, different’ occurs in the main clause. In this predicate lies an ambiguity, as it can either be used as an adjectival or as an adverbial predicate, and the phrase denoting the standard (something to which something else is different) can either be introduced by ocus or be a prepositional phrase with fri. In (21) doḟoirṅde clearly has a direct relative clause marker (i.e. lenition) and is translated with a headless wh-relative clause both in Thes. and in Bauer etal.29 (22) is probably another genuine instance, but it could also be seen in two ways, similarly to (23). In fact (23) can be interpreted as having two conjoined subjects with two slightly different structures (‘A and B are different’), but also as having an adverbial predicate and cleft structure (so Thes. ‘differently is it shown sensibus and as it is afterwards’), therefore without assuming a headless relative clause in donadbantar sensibus. 21) is sain doḟoirṅde són fri-sin roithnigud cop.prs.3sg different.nom.n rel\signify.prs.3sg that against-det.acc.sg.mshining.acc ‘what that [scil. Λευκοθέα] signifies is different from the radiance’ (Sg 36a3) 27. Murphy, 1952, p.149, does not consider bebais a relative form but a Middle Irish form for bebaid (footnote 11). 28. Text ID in CorPH S0003-29. 29. Bauer etal., 2017. 122 ELISA ROMA ETUDES_CELTIQUES_49_CC22_PC.indd 122ETUDES_CELTIQUES_49_CC22_PC.indd 122 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions 22) is sain fu[a] –n-tuc side cop.prs.3sg different.nom.nunder.rel-obj.3sg.m.rel-apply.prF.3sg30 3sg.nom.m fri-s[a]n-í fua-nd-ragab petar against-det.acc.sg.n-deict under.rel-obj.3sg.m.rel-utter.prF.3sg Peter.nom ‘That of which he has applied it is different from that in which Peter has uttered it’ (Ml 38c5) 23) is sain do<n>adbantar sensibus 7 cop.prs.3sg different.nom.n<rel>show.prs.pass.sg [lat]sense.pl.dat and amal ṁbís iarum as rel\be.prs.hab.3sg.rel after.3sg.n ‘what is shown in the senses and how it appears afterwards are different’ (Karlsruhe Augustine, Thes. II, 4, 10a2) In (24) below, the same constructions of the verbs do·beir ‘apply’ and gaibid ‘utter’ with fo ‘about, in relation to’ occur as in (22). The antecedent of the relative clause fuandtuic apstal ‘to which the apostle applied it’ could be the neuter pronoun in the inflected preposition fou; this seems to be the interpretation in the translation by Griffith31 and in CorPH, i.e. ‘for it is not with reference to that that the prophet uttered it: in accordance to which the apostle has applied it, says the commentator.’ However, that antecedent is somewhat far from the relative clause, a situation which is unusual for relative clause heads in Old Irish, unless another construction requires that: therefore, it is possible to interpret the text here as a pseudo-cleft sentence, where the phrase fuandtuic apstal ‘(that) to which the apostle applied it’, a headless relative clause, is cataphorically anticipated by the prepositional phrase fou, just as sentence-final focussed Noun Phrases are anticipated by ed in Old Irish pseudo-cleft sentences (see (6) above). The very same cleft structure occurs at Ml 38c3 ní fou dauc int apstal fon cheill fuandrogab in faith ‘to what the apostle has applied it is not to the sense to which the prophet applied it’, where however there is a nominal antecedent (cheill) in the final focussed phrase with relative clause.32 The translation below tries to pursue this analysis and is closer to the one in Thes., which however avoids the cleft structure (‘for the prophet did not utter it with respect to that to which the apostle applies it, says the commentator’). 24) ar ní fou r<a>gab in faith because neg.cop.prs.3sg under.3sg.n<obj.3sg.n>utter.prF.3sg det.nom.sg.mprophet.nom fua-nd-tuic apstal ol stoirier under.rel-obj.3sg.n.rel-apply.prF.3sg apostle.nom says commentator.nom ‘for it is not with reference to that about which the apostle has applied it that the prophet had uttered it, says the commentator’ (Ml 50d18) 30. That the object pronouns in fuantuc and fuandragab are masculine and not neuter, as suggested by GriFFith and StiFter, 2013, and CorPH, presumably resuming and agreeing with in testimin so in Ml 38c3, is doubtful in my opinion. 31. GriFFith and StiFter, 2013. 32. The lack of agreement between the pronoun in fou (masculine or neuter) and cheill (a feminine noun) makes it clear that the clefted constituent fou anticipates cataphorically here the whole prepositional phrase foncheill fuandrogab in faith, while there is no direct phoric relationship between the pronoun and the noun. HEADLESS RELATIVE CLAUSES IN EARLY IRISH 123 ETUDES_CELTIQUES_49_CC22_PC.indd 123ETUDES_CELTIQUES_49_CC22_PC.indd 123 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions 31) lase dufuinchidet ind-hí when rel\depart.prs.3pl54 det.nom.pl.m-deict dano quos solet experiri aduersitas then [lat]pron.rel.acc.pl.m[lat]is_wont.prs.3sg [lat]prove.inF [lat]adversity.nom ‘when those, then, whom adversity usually tests depart’ (Ml 108b4) 32) hit hé do<d>mainetar insin cop.prs.3pl 3pl <obj.3sg.n.rel>think.prs.3pl that ind-í qui reliqua det.nom.pl.m-deict [lat]pron.rel.nom.pl.m[lat]etcetera ‘it is they who think that, those whoetc.’ (Sg 5a6) 33) conid in spirut adamre so_that.cop.prs.3sg det.nom.sg.mspirit.nom wondrous.nom tra profetauit post passionem then [lat]prophesy.prF.3sg [lat]after [lat]passion.acc ‘so that it is the marvelous spirit that prophesied after the passion’ (Armagh Glosses, Thes. I, 495, 170b1) V. Conclusions and prospects Although the overview presented in this paper would likely benefit from a complement survey of headless relative clauses in the Würzburg Glosses, which as yet do not enjoy the ease of sifting allowed by an annotated database,55 the data gathered through CorPH are large enough to show that headless relative clauses are quantitatively marginal but well-established constructions in Early Irish grammar. Even petrified periphrases appear to stem from headless relative constructions (see (27) above). Their occurrence is rather unrestricted from the syntactic point of view, as the only barred syntactic roles in the corpus are prepositional complement in the matrix clause and genitive in either clauses. Conditions in the relationship between syntactic role in the 54. The meaning of this verb should rather be ‘depart, separate’ (as in Thes.) than ‘descend’ (as in GriFFith and StiFter, 2013 and CorPH, with a question mark), as its explanation with etirscarid at Ml 67b4 suggests. The context is similar in both loci and in both cases where it occurs the Latin verb it renders is *discendo (discendendo, discendunt), but this should rather be emended to forms of discedo. 55. A preliminary survey has given the following results: 4 sure instances of headless relative clauses are attested in the Würzburg Glosses, where the syntactic role in the relative clause is mainly subject and once object, and in the main clause only subject: 18a2 nicon fil bas sciith lim actrop ar crist ‘there is nothing which irks me, provided that it be for Christ’, 23c21 rafitis as lia ‘(what are) most know it’, 32a13 bieid bes ferr de tra ‘there will be something better therefrom then’, 32a22 attá immurgu asbéer ‘there is however something that I will say’. A couple of less sure instances have the same case role in the main and in the relative clause: 8d20 ishé didiu intecttaire maith condaig indocbáil dia thigerni et non sibi ipsi ‘he, then, is the good messenger who seeks glory for his lord and not for himself’, which has cleft-like structure, and 17b6 ni irbágam nádernam iarríchte ní barscéuil si ‘we do not boast what we cannot do after arriving…’, or ‘after tidings of you reach (us)’?). Both animate and inanimate, indefinite specific, non-specific and possibly active referents occur. 130 ELISA ROMA ETUDES_CELTIQUES_49_CC22_PC.indd 130ETUDES_CELTIQUES_49_CC22_PC.indd 130 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions matrix vs. the relative clause rather point to preference for disjoint case marking, i.e. different cases in matrix and relative clause, but same case is also allowed. The preference could be linked rather to the gapping strategy (i.e. absence of relative pronoun and of overt case-marking in the relative clause) than to head omission conditions.56 From the semantic point of view, headless relative clauses in Early Irish can imply animate or inanimate semantics. They can be definite referential, indefinite specific –especially in the universal interpretation with animates– and indefinite nonspecific, similarly to non-finite, participial relative clauses in other languages. To sum up, headless relative clauses in Early Irish appear to be rather free alternatives to headed relative clauses, in particular to open relative clauses with set antecedents such as a n-, intí/aní, cách, nech/ní, nanní. It could even be argued that headless relative clauses are the basic type of relative clauses in Old Irish and that relative clauses with a nominal head can be analysed as nominal expressions in apposition to the head.57 This view cannot be pursued here, but the distribution of definite articles with nouns modified by a relative clause,58 together with the widespread use of relative verb forms in complement clauses (GOI59 and Section 2 above), seems to comply with this analysis. Be that as it may, headless relative clauses in Early Irish are a typologically remarkable feature given that the language does not have any relative pronoun (either whor other). Acknowledgments: I am grateful to Fangzhe Qiu for assisting me in extracting the data from CorPH. All disclaimers apply. Author: Elisa Roma Affiliation: Dipartimento di Studi Umanistici –Università di Pavia e-mail address: [email protected] 56. See Comrie and Kuteva, 2013a, sect. 3. 57. See Dryer, 2013, sect. 1 for this possibility. 58. Roma, 2009, 2014. 59. Thurneysen, 1946, p.318. HEADLESS RELATIVE CLAUSES IN EARLY IRISH 131 ETUDES_CELTIQUES_49_CC22_PC.indd 131ETUDES_CELTIQUES_49_CC22_PC.indd 131 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions REFERENCES Bauer etal., 2017: Bernhard Bauer, Rijcklof HoFman, Pádraic Moran, St Gall Priscian Glosses, version 2.0, 2017, online at http://www.stgallpriscian.ie/. Bisagni, 2014: Jacopo Bisagni, « Prolegomena to the study of code-switching in the Old Irish Glosses », Peritia, 24-25, p.1-58. Breatnach, 1990: Liam Breatnach, « On the citation of words and a use of the neuter article in Old Irish », Ériu, 41, p.95-101. Caponigro, 2021: Ivano Caponigro, « Introducing Headless Relative Clauses and the findings from Mesoamerican languages », in Ivano Caponigro, Harold Torrence, Roberto Zavala Maldonado (eds.), Headless Relative Clauses in Mesoamerican Languages, Oxford, Oxford University Press, 2021, p.1-57. Citko, 2004: Barbara Citko, « On Headed, Headless, and Light-Headed Relatives », Natural Language & Linguistic Theory, 22, 1, p.95-126. Comrie and Kuteva, 2013a: Bernard Comrie, Tania Kuteva, « Relativization on Subjects », in Matthew S. Dryer, Martin Haspelmath (eds.), The World Atlas of Language Structures Online, Leipzig, 2013, online at https://wals.info/chapter/122. Comrie and Kuteva, 2013b: Bernard Comrie, Tania Kuteva, « Relativization Strategies », dans Matthew S. Dryer, Martin Haspelmath (éd.), The World Atlas of Language Structures Online, Leipzig, 2013, online at https://wals.info/chapter/s8. Dryer, 2013: Matthew S. Dryer, « Order of Relative Clause and Noun », in Matthew S. Dryer, Martin Haspelmath (eds.), The World Atlas of Language Structures Online, Leipzig, 2013, online at https://wals.info/chapter/90. eDIL, 2019: eDIL, « An Electronic Dictionary of the Irish Language, based on the Contributions to a Dictionary of the Irish Language (Dublin: Royal Irish Academy, 1913-1976) », 2019, online at https://dil.ie/. García Castillero, 2018: Carlos García Castillero, « The name and its circumstances: On the recognitional use of the Old Irish demonstrative intí », in José M. Vallejo, Iván Igartua, Carlos García Castillero (eds.), Studia Philologica et Diachronica in Honorem Joaquín Gorrochategui. Indoeuropaea et Palaeohispanica, Anejos de Veleia Series minor 35, Vitoria-Gasteiz, 2018, p.147-169. García Castillero, 2023: Carlos García Castillero, « On the NP syntax in the Old Irish glosses: The missing data of the isolated light-headed NPs », paper presented at the Workshop Nominal Syntax in the Medieval and Early Modern Celtic Languages, Göttingen, 27-28 October 2023. GriFFith and StiFter, 2013: Aaron GriFFith, David StiFter, A dictionary of the Old-Irish glosses in the Milan Codex Ambrosianus C 301 inf, 2013, online at https://indogermanistik. univie.ac.at/milan-glosses/. Haspelmath, 1997: Martin Haspelmath, Indefinite Pronouns, Oxford, Oxford University Press, 1997. Haspelmath, 2013: Martin Haspelmath, « Indefinite Pronouns », in Matthew S. Dryer, Martin Haspelmath (eds.), The World Atlas of Language Structures Online, Leipzig, 2013, online at https://wals.info/chapter/46. 132 ELISA ROMA ETUDES_CELTIQUES_49_CC22_PC.indd 132ETUDES_CELTIQUES_49_CC22_PC.indd 132 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions Murphy, 1952: Gerard Murphy, « On the dates of two sources used in Thurneysen’s Heldensage », Ériu, 16, p.145-156. Ó Cathasaigh, 1990: Tomás Ó Cathasaigh, « On the Early-Irish prepositional relative without antecedent », Celtica, 21, p.419-426. Pedersen, 1909: Holger Pedersen, Vergleichende Grammatik der keltischen Sprachen, 2 vols., Göttingen, Vandenhoeck und Ruprecht, 1909. Roma, 2007: Elisa Roma, « Relativisation Strategies in Insular Celtic Languages: History and Contacts with English », in Paolo Ramat, Elisa Roma (eds.), Europe and the Mediterranean as Linguistic Areas. Convergencies from a Historical and Typological Perspective, Amsterdam, Benjamins, 2007 (Studies in Language Companion Series), p.245-288. Roma, 2009: Elisa Roma, « How many definiteness markers per NP in Old Irish? », in Stefan Zimmer (ed.), Kelten am Rhein: Akten des dreizehnten Internationalen Keltologiekongresses, 23. bis 27. Juli 2007 in Bonn, Mainz, Philipp von Zabern, 2009, p.223-231. Roma, 2014: Elisa Roma, « Old Irish noun phrases: data from the Milan Glosses and a hypothesis for the origin of the single article constraint », in Elisa Roma, David StiFter (eds.), Linguistic and philological studies in Early Irish, Lewinston, NY, Edwin Mellen, 2014, p.131-176. Sadler and Camilleri, 2018: Louisa Sadler and Maris Camilleri, « Free relatives in Maltese », Brill’s Journal of Afroasiatic Languages and Linguistics, 10, p.115-159. StiFter et alii, 2021: David StiFter, Bernhard Bauer, Elliott Lash, Fangzhe Qiu, Nora White, Siobhán Barrett, Aaron GriFFith, Romanas Bulatovas, Francesco Felici, Ellen Ganly, Truc Ha Nguyen, Lars Nooij, Corpus PalaeoHibernicum (CorPH) v1.0, 2021, online at https://chronhib.maynoothuniversity.ie/chronhibWebsite/home. Stokes and Strachan, 1901 (Thes.): Whitley Stokes, John Strachan, Thesaurus palaeohibernicus, 2 vols., Cambridge, Cambridge University Press, 1901. Thurneysen, 1946 (GOI): Rudolf Thurneysen, A grammar of Old Irish, Translated from the German by D. A. Binchy and O. Bergin, Dublin, Dublin Institute for Advanced Studies, 1946. HEADLESS RELATIVE CLAUSES IN EARLY IRISH 133 ETUDES_CELTIQUES_49_CC22_PC.indd 133ETUDES_CELTIQUES_49_CC22_PC.indd 133 30/01/2024 17:5430/01/2024 17:54 © CNRS Éditions