Morpheme-internal Recursion in Phonology

Kuniya Nasukawa, Kuniya Nasukawa

• Publisher: De Gruyter Mouton
• Year: 2020
• ISBN:
  9781501512414
• Edition: 1
• Topic: Filología
• Subtopic: Sintaxis en lingüística

Generative phonology aims to formalise two distinct aspects of phonological processes: the functional and the representational. Since functions operate on representations, it is clear that the functional aspect is influenced by the form of representations, i.e. different types of representation require different types of rules, principles or constraints. This volume examines the representational issue in phonology and considers what kind of representation is most appropriate for recent models of generative phonology. In particular, it provides the first platform for debate on the place of morpheme-internal structure and on the formal status of phonology in the language faculty, and attempts to identify phonological recursive structure as a means of capturing frequently observed processes.

Obstruent liquid clusters: Locality, projections and percolation

Ali Tifrit

1 Introduction

Rhotics and laterals (Rice and Avery 1991, Rice 1992, Backley 2011) act as targets for the same processes and share distributional properties, one of which is the possibility to form an onset cluster with an obstruent. This is one of the main arguments to group rhotics and laterals together as a natural class (Wiese 2011). In this paper, I investigate the behavior of liquids and the case of /Obstruent+Liquid/ (OL) clusters in a modified Government Phonology 2.0 framework (GP2.0: Pöchtrager 2006, Pöchtrager and Zivanovic 2010).

Most of the recent work on OL clusters have been conducted in a CV framework (Lowenstamm 1996, Scheer 2004). I first illustrate the questions arising in this perspective and, in particular, the cases where OL clusters act sometimes as a single element and sometimes as two distinct objects. I then underline the theoretical issues that are related to the flatness of the CV model (Locality and Infrasegmental Government). I propose that the way out is to consider that the representations contain much more structure than what is generally assumed. After discussing the internal content of liquids, I put forward that contrary to other classes of consonants, what characterizes liquids is their inability to project and their search for a hosting structure. I then reconsider classical cases of lenition, surface changes, compensatory lengthening by loss of an onset and metatheses by formalizing them in GP2.0.

2 Unexpected properties of OL clusters

An interesting part of the activity of the liquids is their ability to contract a relationship with obstruents. There is a debate concerning this affinity: should we consider Obstruent+Liquid as ‘clusters’, i.e. as two independent consonants (potentially separated by an empty nucleus: √OvL)? Or should we analyse OL as a single object: an ‘affricate’ linked to one consonantal position (√OL)? In this part, I review evidence of the peculiar behaviour of OL clusters, mainly in a CV framework (Lowenstamm 1996, Scheer 2004). The question relies primarily on the monopositionality or the bipositionality of these clusters. I discuss the treatment given to these groups in this framework and underline the problems they cause. Next, I examine data from French with the case of the reduction of OL+schwa#, and the cases of intervocalic lenition of OL clusters from Latin to French and in Gorgia Toscana. I conclude this section with the problem of locality.

2.1 Mono or bipositionality

Lowenstamm (2003), Scheer and Ségéral (2007), Scheer (2014) point out, in a CV framework, that OL clusters can be analysed as occupying one or two positions. Both types may coexist in the same language as is the case in Czech, for example. Lowenstamm (2003: 12–15) argues that the two representations can explain epenthesis targeting some prefixes in Czech: the following examples (1a) and (1b) show a contrastive behaviour related to the structure of the OL cluster.

(1)

Czech (Lowenstamm 2003:15)

In (1a), epenthesis occcurs at the end of the inchoative prefix [roz] because of the presence of an unexpressed vowel between the obstruent /d/ and the liquid /r/ in [drat]. This empty nucleus is unable to properly govern the preceding vowel V₀. Hence, the ungoverned V₀ surfaces and we obtain [roze-drat].

However, if the nucleus V₁ is filled with segmental material, as in [deru], it is able to properly govern the preceding nucleus and V₀ remains empty: no epenthesis happens and for example, [roz-deru] surfaces. These examples clearly show the alternation between a full vowel and an empty vowel inside the OL cluster. We can hypot...