Institut de Lingüística Aplicada

Third Mediterranean Meeting on Morphology (MMM3)


Paolo Acquaviva (University College Dublin)
Agreement in autonomous morphology: Evidence form Irish Prepositions

Both verbs and prepositions are inflected in Irish, but two prepositions have a curious property: whenever they linearly precede the article, irrespective of constituency, they must appear in the [3 sg masc] form. For example, le 'with' is unchanged when it's followed by bean 'woman': le bean eile 'with another woman', but it must appear as leis (lit. 'with.him/it') if the complement noun is introduced by the article: leis an bhean 'with the woman'. The same applies to trí 'through'. This is puzzling: first, because the masculine feature of the P may well clash with the gender of the complement DP (as happens with the feminine an bhean); secondly, nowhere else in the Irish conjugation can an agreement suffix appear alongside a full DP (suffixes and overt DPs are notoriously mutually exclusive in Modern Irish); thirdly, it is unusual that "agreement" should depend on linear, rather than hierarchical, relations.

This type of "parasitic agreement" would appear to instantiate a rule of referral: le and trí take the [3 sg masc] form when followed by the article. But a more revealing analysis can be offered.

The values [3 sg masc] are clearly a default; yet the preposition also has an uninflected base form (the citation form), used when the article is absent; e.g. le Nuala 'with Nuala'. Because of the existence of this uninflected form, leis cannot be regarded as the elsewhere case for the whole paradigm; it is only the default among the inflected forms. If we want to keep it as a default form, which does not explicitly specify the features [3 sg masc], there must be a way to represent an inflected form without specifying any pronominal features.

I propose that the stem of Irish prepositions contains an inherent feature AGR, corresponding to the category of pronominal agreement, but unspecified for any one value (person, gender and number are always fused together in the Irish conjugation). A first result follows immediately: since the feature is lexically related, we expect to find exceptions; and indeed a few Irish prepositions cannot inflect (unlike verbs).

If the nominal argument contains pronominal features other than [3 sg masc], and nothing else (i.e. if it's a simple pronoun), these features provide a value for the AGR feature of the preposition, which is spelt out as a suffix. The suffix, thus, is just the realization of a pronominal argument; that's why suffixed forms behave syntactically as if they governed a pronoun. The combination [3 sg masc], however, is the default value for AGR; it is realized by a special stem form, not by a separate suffix. To my knowledge, it had never been noted that this is the only form for which there is no single suffix for all prepositions.

If the argument is a full DP, it is this DP that provides exponence for AGR - along with the additional grammatical information contained. This analysis makes precise the intuition that agreement suffixes and full DP complements are alternative realizations of the same content; in Irish, unlike in other languages, no "agreement" is involved between nominals and verbs/prepositions. A form like leis is thus analyzed as WITH+AGR, with no values for the latter category. Since this is not, formally, a [3 sg masc] form, it is easier to understand how it can be employed in contexts like leis an bhean, with a fem. DP. In fact, the AGR of the preposition here has nothing to do with the following DP. I propose that AGR is here inserted between the preposition and the article, as an instance of morphological readjustment in front of the article - a very common occurrence in the Irish morphonology (for example, the preposition de 'of' and the plural article na are realized as desna in some dialects, with an added s). The case of leis is unusual because the inserted element is not segmental material but an abstract morphological slot: the string of morphological units P+ARTICLE becomes P+AGR+ARTICLE, for two lexically determined prepositions. What used to be a morphonological alternation in Old Irish has now become a purely morphological juncture phenomenon. Note that the parallelism with other epenthesis phenomena would be lost if this was just a rule of referral.

The hypothesis of an epenthetic AGR slot explains why the two prepositions have the [3 sg masc] form; this is the default inflection. Only in this case does the suffix coexist with an argument DP, because only here is AGR added postsyntactically. For the same reason linear order is relevant, and not syntactic constituency. This account is based on the notion of a purely morphological AGR characterization that combines with verbs and prepositions, an abstract concept that mediates in a complex way between syntactic features and phonological exponence. No such analysis can be proposed without recognizing a morphological module between syntax and phonology. The insertion of an extra AGR in a particular context is but a minor, but revealing, side-effect.