Noam Faust (SFL), Francesc Torres-Tamarit (SFL) and Shanti Ulfsbjorninn (Lyon III and UCL)

17
fév.
2016.
10h00
13h00

Pouchet, salle 159

Three talks on stress in CVCV

Talk 1
Noam Faust
 (SFL) and Shanti Ulfsbjorninn (Lyon III and UCL)
Arabic stress with no moras, no extrametricality and no syllables

Abstract - Following Scheer & Szigetvary (2005) and Ulfsbjorninn (2014), this talk models stress assignment using empty nuclei. We consider two dialects of Arabic, Palestinian (PA) and Cairene (CA). Both have: 1) antepenultimate stress in LLL words [kátabu] ‘they wrote’ and 2) stress-to-weight otherwise [katáb-t] ‘I wrote’ [biħíbb] ‘he shall love'. They differ in HLL cases: PA has [sákkatat] ‘she silenced’, CA has [sakkátat]. PA additionally has stress shift to the antepenultimate if the heavy syllable becomes preantepenultimate: [sakkátato] ‘she silenced him’. The standard account (Hayes 1995) uses both moras and extrametricality, which are both stipulations whose only motivation comes from getting the stress facts right. The present account covers the same facts without using moras or extrametricality, and the tools it does use (EN, FEN and government) are all independently motivated. Thus, the present account fares better on both economy and elegance.

Talk 2
Shanti Ulfsbjorninn (Lyon III and UCL)
Oceanic Diphthongs and their relatives: a Strict CV analysis

Abstract - Oceanic stress typology demonstrates the benefit of using a Strict CV grid theory in the analysis of metrical phenomena and highlights the representational difference between hiatus (VV sequences) and diphthongs (VVs in a dependency relationship). Scheer and Szigetvari (2005)'s contribution to metrical theory was the proposed unification of syllabic and metrical structure; this works by building a grid of metrical projections directly on a strictly alternating CV skeleton. One of the key advantages of this model is the ability to reveal something hinted at in work by Burzio (1991) and Harris and Gussmann (2002): many quantity driven stress systems are better analysed as fixed stress systems when empty nuclei are also projected onto the grid. The suggestion of this analysis is that perhaps all quantity could be dispensed with, in line with Strict CV's flat, constituent-less representations. Ulfsbjorninn (2014) showed, however, that some languages (Pulaar, Wolof, Kashmiri) have true quantity-determined stress. Even when analysed in a Strict CV grid theory, Pulaar (for example) required a grouping function to be added to the model. Therefore, in a select few languages one could talk of CVVC, CVV and CVC units. This naturally raises the question of the representation of diphthongs, an area so far understudied in Strict CV. In this talk we will present four language types (common in the region). The first, the Vinmavis type, has penultimate stress in both vowel-final and consonant-final words and is traditionally seen as a fixed stress language. The second, the Tamambo type has penultimate stress on vowel-final words, but final stress on consonant-final words - traditionally it is analysed as quantity driven. In both language types however, VV sequences are identically stressed; 'VV when VVs are found at the end of the word. Under the standard approach, this same behaviour of VVs has to be analysed very differently in each language type. In the Vinmavis type, one must propose that feet are built on moras. While in the Tamambo type, one assumes quantity sensitivity of VV sequences. Neither analysis is complicated, but it misses the primary observation: VVs have the same exact behaviour in the two language types. VVs in these languages are hiatuses not diphthongs. In Strict CV grid theory, Vinmavis and Tamambo languages have fixed penultimate stress in all words. The only difference between the languages is whether empty nuclei are metrically projected. We then move to the third language type (Sye), this one is interesting because it appears to have the same penultimate fixed stress system as the Vinmavis and Tamambo types, but unlike these it appears to count VVs as a single unit - real diphthongs. However, simple syllable structure analysis shows that this language does not have final VV sequences, it has final VC sequences and, therefore, is identical to the previously described Vinmavis type. The fourth language type we present, represented by Island Tigak does have true diphthongs, and is analysed as a true quantity determined stress system. With this kind of evidence, we will show three structural descriptions for what are traditionally called diphthongs: hiatus (VV sequences), bogus diphthongs VC sequences, and real diphthongs (VVs in a dependency relationship).

Talk 3
Noam Faust (SFL) and Francesc Torres-Tamarit (SFL)
Identifying positions, or why final /n/ deletion in Catalan is not so crazy

Abstract - In Catalan /n/ deletes when word-final and following a stressed vowel, as in comú 'common', but comuns 'common.pl.', comunitat 'comunity'; ple 'full', but plens 'full.pl.', pleníssim 'very full' (boldfaced syllables indicate stress). That stress is an obligatory condition is shown in cases like plàtan 'banana' or origen 'origin', where final /n/ fails to delete in words without final stress. Nevertheless, one does find final nasals preceded by stress. These exceptions occur under three conditions: (i) if [n] is the second consonant of a word-final cluster, as in carn 'meat' or hivern 'winter'; (ii) if [n] is the first consonant in an underlying final /nt/ cluster that has been simplified by deleting the stop, as in [dén] 'tooth' (cf. dental 'dental'); (iii) in a list of lexical exceptions within the native vocabulary of the language, as in son 'sleep, sleepiness', nen 'kid', arran 'close'; and (iv) in many loan and learned words, as in magazín 'magazine', orangutan 'orangutan', xarleston 'charleston', although substandard forms in which the final /n/ deletes are also found (magazí, orangutà, xarlestó), as will be discussed. We will present an analysis of (blocking of) /n/ deletion in Catalan that will make use of Strict CV representations and violable constraints in Optimality Theory, two theories that are rarely combined, but completely compatible. We will propose that final /n/ floats unassociated below its C position if it cannot be licensed from a following segment. Word-finally, this is not possible because there is no licensor at the right of /n/. This non-association of /n/ may create the problem of unidentified CV structure. However, we will see that when stress is final, the segmentally unidentified final CV is nevertheless engaged by the metrical structure---the foot's dependent coincides with the final empty nucleus. If so, only when stress is not final does that final CV structure remain unidentified. And only in those cases the /n/ must be associated to its C position even though /n/ is unlicensed. The analysis will also be extended to the rest of the data in which final /n/ deletion fails to apply and to lexical exceptions in the native vocabulary. We will also discuss the issue of process productivity in the light of loan and learned words showing variation with respect to final /n/ deletion.