Verbal Morphology and Clausal Projections in Early Irish

Cathal Doherty
University College Dublin

Early Irish verbal morphology is notoriously complex in that verbs distinguish two sets of endings for most tenses and moods, the use of which varies with syntactic context: i.e. the 'absolute' and 'conjunct' paradigms (illustrated below for beirid Œcarry¹ in the present indicative):

(1)     Absolute      Conjunct          Absolute     Conjunct
Sg. 1.  biru          ·biur     Pl. 1.  berm(a)i     ·beram
    2.  biri          ·bir          2.  beirthe      ·berid -ith
    3.  berid -ith    ·beir         3.  ber(a)it     ·berat

This paper argues that the distribution of these endings reflects clause structure, specifically whether CP is projected:

(2) Conjunct endings appear when CP is projected, while absolute endings appear in minimal clauses projecting only to IP.

This proposal is supported not only by clauses in which Cš is overt, but also by a variety of apparently disparate sentence-types, including imperatives and (residual) Verb Second sentences. Traditional grammars report that conjunct forms typically appear after the class of 'conjunct particles':

(3)	ní		NEG				
	co / con	so that
	dia		if
	ara		in order that           (GOI 28)

As these particles are presentential, stressless proclitics which introduce different clause-types, it is clear that they belong to the complementizer class. The use of the conjunct paradigm after these particles, as in (4) below’ therefore straightforwardly follows from (2) above:

(4)a.	beirid 'he carries'	b. ní·beir 'he does not carry'

Imperative sentences also provide further evidence for (2). Imperatives in Old Irish are not introduced by a preverbal particle and distinguish only one set of endings, which are almost identical to the conjunct endings in (1) above. Assuming that CP is projected in imperatives (e.g. following Potsdam 1996), these facts follow from the generalization in (2). Old Irish makes extensive use of compound verbs, consisting of a verbal root compounded with one or more prepositional preverbs (which radically and unpredictably change the lexical meaning):

(5)	gairid		Œcalls¹	       V
	do·gair 	Œsummons¹    P·VCONJ	(*do·gairidABS)
	for·cum-gair 	Œcommands¹   P·PVCONJ	(*for·cum-gairidABS)
	do·air-n-gair 	Œpromises¹   P·PPVCONJ	(*do·air-n-gairidABS)

It is remarkable that the verbal root in these compounds exclusively takes the conjunct endings, as illustrated above. These facts indicate that compound verbs necessarily involve the projection of CP: in particular, that the preverbs are associated with the CP projection. There is strong evidence for this position from the behavior of preverbs in embedded contexts. Old Irish lacks a subordinating complementizer, but often uses initial mutation (nasalization) as a marker of subordination in complement and relative clauses. It is remarkable that in the case of an embedded compound verb, subordinating nasalization appears to the right of the initial preverb:

(6)     epert    frissom  rondbiad     fáilte libsi       (Wb. 16b19)
        say.INF  to.him  joy
       ŒŠ to say to him that there will be joy with you.¹

Assuming that subordinating nasalization is associated with the CP projection, this fact indicates the presence of (at least the leftmost) preverb in CP. The obligatory use of conjunct endings in compound verbs then follows straightforwardly from (2). Finally, although Old Irish is generally a strict VSO language (as is Modern Irish), many residues of an earlier V2 stage remain, e.g. the archaism known traditionally as 'Bergin's construction' (Bergin 1938, Doherty 1997):

(7)a. [ lÇithe  Gailïoin ]   gabsat               inna     lÇmaib  lÇigne
      warriors  Gailïon.GEN  gabaid.3PL.PRET.CONJ in.their hands.DAT  spears
       ŒThe Galïon warriors took spears in their hands.¹ 

   b. [ BÇngluinn ]      gn¥                glenn       gaeth.
      bloodless.deed  gníid.3SG.PRES.CONJ  valleys.GEN  wind
       ŒThe wind of the valleys does a bloodless deed¹ 
   c. [ srethaib  sluag ] soí                   Crimthan Coscrach   cing   cét   catha
       lines.DAT  hosts  sóïd.3SG.PRET.CONJ C.  victorious champion  100 battles
       ŒWith lines of hosts’ C. the Victorious the champion won (turned) a hundred battles.¹

These clauses are remarkable in that the verb appears in the conjunct form, although it is not preceded by an overt complementizer. Under the generalization in (2), however, these facts follow straightforwardly, as the standard analysis of V2 involves the projection of CP. In later language, absolute endings are found in V2 sentences, (8) below, however:

(8)  Láir  dano    bói              i  ndorus in tige     trogais            da lurchuire
     mare moreover be.3SG.REL.PRET  in door   the house trogaid.3SG.PRET.ABS two foals
     ŒA mare which was before the house dropped two foals¹  

This follows assuming that the V2 position shifted to Spec of IP. Note that Spec of IP is an A´-position under the analysis of Irish clause structure advanced in McCloskey's work (e.g. McCloskey 1996). Therefore, V2 sentences in a VSO language provide no evidence for the projection of CP, as Spec IP is available as the V2 Topic position. A process of language change whereby V2 sentences were reanalyzed as IP accounts straightforwardly for the use of absolute endings in these later V2 sentences. It is also a strong prediction of this account that embedded V2 should be possible, and remarkably, this is borne out: examples of embedded V2 do exist in Old Irish. In sum, the proposal that the conjunct paradigm of Old Irish verbs indicates the projection of CP has considerable evidence in its favor and provides the basis for an explanation of a broad range of seemingly unrelated facts.


Begin, Osborn. 1938. On the Syntax of the Verb in Old Irish. Ériu 12: 197-214.
Doherty, Cathal. 1998. Residual Verb Second in the Syntax of Early Irish. 18th Harvard Celtic Colloquium.
McCloskey, James. 1996. Verb Movement in Irish. Natural Language and Linguistic Theory 14.1
Potsdam, Eric. 1996. Syntactic Issues in the English Imperative. Ph.D. dissertation, University of California, Santa Cruz.

< Back to List of Abstracts | Back to PLC23 Home Page >

About the PLC23 Committee
Previously held Penn Linguistics Colloquium: PLC22 (1998), PLC21 (1997)

Penn Department of Linguistics
University of Pennsylvania