Phrasal Affixation and the Syntax/Morphology Interface

It has become a commonplace to argue that pronominal clitics in Romance languages have many of the properties of affixes rather than clitics (Halpern 1995, Miller & Sag 1997, Monachesi 1999). This is particularly clear in the case of European Portuguese enclitics, which pass all the Pullum-Zwicky (1983) tests for affixes: they form a fixed cluster in which the clitics show idiosyncratic allomorphy and they invariably attach directly to a verb (1,2). Moreover, in the Future/Conditional the pronominal clitics appear between the verb stem and the tense/agreement endings (so-called 'mesoclisis', a phenomenon which has thus far eluded insightful description, see (3,4)). Finally, they trigger idiosyncratic stem allomorphy, a crucial sign of affixation (5,6).

However, under certain syntactic and semantic conditions ('proclisis triggers') exactly the same cluster exhibits proclisis (7). Here, the attachment to the verb is much less strict and the cluster shows the properties of phrasal affixation. The proclitics are left adjacent to the edge of the VP and can be separated from the lexical or auxiliary verb only by a handful of (mainly monosyllabic) adverbs. However, the cluster can have scope over conjoined VPs (which is impossible with the enclitics) (see (8,9)).

Crysmann (1997,2001) has recently proposed an HPSG analysis which essentially "liberates" the morphological clusters into the syntax for linearization. We concur with his general thesis but note several inadequacies in his approach. In particular, we argue that the enclitics are straightforwardly inflections attaching to stems, and hence placed morphologically, while the proclitics are phrasal affixes whose placement must make appeal to syntactic (phrasal) categories.

The enclitic and proclitic clusters are absolutely identical so we have the rarely-discussed circumstance in which stem-based affixation and phrasal affixation places the same set of objects. This is a phenomenon which has been largely ignored in the syntactically-oriented literature ('suspended affixation' in Turkish is a partial exception), and we provide further exemplification of this phenomenon, showing that it is not as rare as the paucity of discussion would suggest. This raises an interesting problem, however, both for theories of affixation and for principles such as lexical integrity which requires the strict separation of syntax and morphology, and it is these aspects we focus on in this paper.

We propose a modest extension to the inferential-realizational Paradigm Function Morphology of Stump (2001) to accommodate the flexible placement of affixes such as the EP pronominal object inflections (cf. also Stump 1993 on Swahili ambifixation). Affixes themselves are pure form. The format of the realizational rule is then split into components, exponence (specifying which affix is added as the realization of the rule's feature set) and placement (left/right alignment). The realization rules are then reformulated so as to generate affix/clitic clusters in the absence of a specific stem/host, without altering any other aspects of Stump's machinery. We show that this separation of exponence from placement provides for a very simple description of mesoclisis and also makes available a clitic cluster which can be placed phrasally.

With this morphological machinery in place, we examine in more detail the status of phrasal affixation in general and in particular the light that this phenomenon sheds on the nature of the relation between syntactic structure and its morphological expression. We argue that phrasal affixes such as those found in EP procliticisation contribute their f-structure information to phrasal nodes and are not represented as terminal elements in the syntactic tree. The asymmetry between the scopal properties of the enclitic and the proclitic falls out from the phrasal affixation analysis in conjunction with the LFG theory of coordination. The interpolation of X0 elements between phrasal affix and head follows from the statement of placement requirements in the morphology. We examine the apparently unrelated set of syntactic configurations and constructions which trigger procliticisation in EP and propose that an abstract syntactic feature is responsible for triggering the phrasal affixation placement principle is invoked (see Sells 2001 for a similar use of an abstract syntactic feature in triggering special morphology in Swedish). Elsewhere, the 'normal' morphology supplies the appropriate suffixed verb forms (enclisis). We offer a number of arguments to show that enclisis is the elsewhere condition.

References

• Berthold Crysmann (1997). Cliticization in European Portuguese using parallel morpho-syntactic constraints, Proceedings of LFG97.
• Berthold Crysmann (2001). Clitics and Coordination in Linear Structure, in B. Gerlach et al (eds) Clitics in Phonology, Morphology and Syntax, John Benjamins, Amsterdam.
• Aaron Halpern (1995). On the Placement and Morphology of Clitics, CSLI Publications, Stanford, CA.
• Philip Miller and Ivan Sag (1997). French clitic movement without clitics or movement, Natural Language and Linguistic Theory, 15/3:573-639.
• Paula Monachesi (1999). A lexical approach to Italian cliticization, CSLI Publications, Stanford, CA.
• Greg Stump (2001). Inflectional Morphology, CUP, Cambridge.
• Peter Sells (2001). Syntactic Information and its Morphological Expression, Manuscript, Stanford.

Not submitted; contact authors for original LFG02 slides or current manuscript.

Related work includes:

• Luis and Sadler (to appear 2002). Object Clitics and Marked Morphology. In C. Beyssade, O. Bonami, P. Cabredo Hofherr and F. Corblin Eds Empirical Issues in Formal Syntax and Semantics 4 Presses Universitaires de Paris Sorbonne.
  Available via Sadler's web page.