Vowel length in Icelandic Vowel length in Icelandic compounds and - - PowerPoint PPT Presentation

Vowel length in Icelandic Vowel length in Icelandic compounds and the role of FENs compounds and the role of FENs

Marcin Fortuna Marcin Fortuna Ludwig Ludwig-

• Maximilians

Maximilians-

• Universit

Universitä ät M t Mü ünchen nchen marcin.fortuna@lipp.lmu.de marcin.fortuna@lipp.lmu.de

Government Phonology Round Table 9 Budapest, 20.04.2013

# 2 Government Phonology Round Table 9, Budapest, 20 April 2013

The purpose of this paper: to discuss the phenomenon of ‘post-lexical’ syllabification in Icelandic to argue for a modified model of Strict CV which eliminates Proper Government to elaborate on the mechanism of Direct Interface

# 3 Government Phonology Round Table 9, Budapest, 20 April 2013

Introduction Icelandic: a language with tonic lengthening all stressed vowels in open syllables are long:

bú [pu:] ‘farm’ búa [ˈpu:a] ‘to live’ taka [ˈtha:kha] ‘to take’ sötra [ˈsø:thra] ‘to slurp’ götva [ˈkø:thva] ‘to discover

word-final consonants are extrametrical

þak [θa:kh] ‘roof’, hús [hu:s] ‘house’, vor [vɔ:r] ‘spring’

# 4 Government Phonology Round Table 9, Budapest, 20 April 2013

The last group: the object of the present study CVC words display untypical behaviour when they are the first member of a compound What can happen with the vowel in a compound? Intuitively:

• If the second member is concatenated analytically, the vowel is

lengthened

• If the second member is concatenated synthetically, the vowel is

in a closed syllable and hence resists lengthening

# 5 Government Phonology Round Table 9, Budapest, 20 April 2013

This should be decided by the morphosyntax and be a property of a given morpheme Morphemes induce spell-out or not Sometimes: semantic factors play a role (e.g. whether the meaning is compositional) But… spell-out cannot be melodically conditioned

# 6 Government Phonology Round Table 9, Budapest, 20 April 2013

Scheer (2012b, 2013): melody-free syntax there is some communication between morphosyntax and prosody morphosyntactic computation cannot make reference to melodic primes this violates modularity however… melodic conditioning for spell-out is what Icelandic seems to display

# 7 Government Phonology Round Table 9, Budapest, 20 April 2013

1) The data Data set #1: CVC words ending in {H}-less segments von [vɔ:n] ‘hope’ von+legur [ˈvɔnlɛɣʏr] ‘reliable’ von+laus [ˈvɔnlœʏs] ‘hopeless’ haf [ha:v] ‘ocean’ haf+kola [ˈhavkɔla] ‘sea breeze’ vor [vɔ:r] ‘spring’ vor+kuldi [ˈvɔr̥kʏltɪ] ‘spring chill’ rauður [ˈrœʏ:ðʏr] ‘red’ rauð+leitur [ˈrœʏðlɛithʏr] ‘reddish’

# 8 Government Phonology Round Table 9, Budapest, 20 April 2013

Data set #2: CVC words ending in segments contaning {H} bak [pa:kh] ‘back’ bak+poki [ˈpa:khphɔchi] ‘rucksack’ hvítur [ˈkhvi:thʏr] ‘white’ hvít+leitur [ˈkhvi:thlɛithʏr] ‘whitish’ brosa [ˈprɔ:sa] ‘to smile’ bros+legur [ˈprɔ:slɛɣʏr] ‘smiling’ kátur [ˈkhau:thʏr] ‘merry’ kát+legur [ˈkhau:thlɛɣʏr] ‘funny’

# 9 Government Phonology Round Table 9, Budapest, 20 April 2013

An empirically adequate generalisation:

• If the last consonant of the first member of the compound ends in a

fortis plosive or /s/ (=contains {H}), then concatenate the second member analytically

• If the last consonant of the first member of the compound is any
• ther consonant (=does not contain {H}), then concatenate the

second member synthetically

In other words: spell-out is melodically conditioned (an unacceptable conclusion)

# 10 Government Phonology Round Table 9, Budapest, 20 April 2013

2) Interpretations Árnason (2011):

• two versions of the rule of lengthening: lexical and post-lexical
• the lexical rule applies:

– morpheme-internally – on boundaries with inflectional endings – on boundaries with some derivational morphemes

• the post-lexical rule applies:

– on boundaries with some other derivational morphemes – in compounds – it may apply between any two adjacent words in a sentence (precise morphosyntactic contexts call for more research)

• no attempt of explanation of the post-lexical syllabification algorithm;
• nly a general description

# 11 Government Phonology Round Table 9, Budapest, 20 April 2013

• older approaches:

– Oresnik (1971): compounds may have a # boundary

• explains the behaviour of clusters with {H} but not of the other

group – Gussmann (1985: 90):

• individual morphemes come with a given type of boundary
• compounds have #, later: boundary weakening after voiced

consonants – Booij (1986: 14) transforms this rule into the one which merges two phonological words into one when the first one ends in a voiced consonant

# 12 Government Phonology Round Table 9, Budapest, 20 April 2013

3) Gussmann’s SGP account Gussmann (2002, 2006b): a syllabification algorithm which makes use of empty nuclei (SGP) Regularity:

• stressed rhymes always need to branch
• aspirated plosives can never be syllabified in the coda
• therefore they land in the onset and the preceding nucleus

branches

# 13 Government Phonology Round Table 9, Budapest, 20 April 2013

rauð+leitur [ˈrœʏðlɛithʏr] ‘reddish’

• [ð] can be syllabified in the coda

# 14 Government Phonology Round Table 9, Budapest, 20 April 2013

hvít+leitur [ˈkhvi:thlɛithʏr] ‘whitish’

• [th] cannot be syllabified in the coda; it lands in the onset; the

preceding nucleus branches

# 15 Government Phonology Round Table 9, Budapest, 20 April 2013

„Within the syllabic approach we need to say nothing in addition to what has already been established, namely that aspirated plosives canonly appear in the onset. When this principle is followed, it is obvious that the preceding syllable is open and its nucleus has to branch. No separate generalisations for simplex and complex words are necessary.” (2002: 183) Gussmann tries to be as representational as possible His proposal works only because he refers to surface aspirated plosives!

# 16 Government Phonology Round Table 9, Budapest, 20 April 2013

There are also underlying fortis plosives in the preconsonantal position On the surface they behave differently when the cluster is domain-internal and when it is an effect of concatenation Across a (strong) morpheme boundary:

• /Vth+l/ > [V:thl] (lengthening, post-aspirated plosive)
• /khvith/ + /lɛithʏr/ → [ˈkhvi:thlɛithʏr]

Morpheme-internally/across a weak boundary:

• /Vthl/ > [Vhtl] (no lengthening, preaspiration)
• ætla ‘intend’ /aithla/ → [ˈaihtla]

# 17 Government Phonology Round Table 9, Budapest, 20 April 2013

If Gussmann’s (2002) syllabification algorithm operated blindly (on underlying plosives), it would not distinguish between hvítleitur and ætla Gussmann fails to recognize the importance of the morpheme boundary in syllabification A syllabification algorithm which does not consider the boundary makes a wrong prediction!

# 18 Government Phonology Round Table 9, Budapest, 20 April 2013

A similar problem: the syllabification of /s/ Gussmann (2002: 187-193) calls it a ‘double agent’ it sometimes syllabifies as a coda, sometimes as an

• nset

no attempt of explanation which way is preferred when Actually:

• coda morpheme-internally: taska [ˈthaska] ‘bag’, veisla [ˈveisla]

‘party’

• onset on the boundary: bros+gjarn [ˈpro:scartn̥] ‘funny’,

bros+legur [ˈprɔ:slɛɣʏr] ‘smiling’

# 19 Government Phonology Round Table 9, Budapest, 20 April 2013

• Acc. to Scheer (2012a: 145): SGP has never had a way to

represent morpheme/word boundary neither syllabic arborescence nor skeleton qualify this fact is evident from the Icelandic data! they are insolvable with SGP tools

# 20 Government Phonology Round Table 9, Budapest, 20 April 2013

How much can Direct Interface (Scheer 2012a) help us? Morpheme boundaries materialise themselves in the representation as empty syllabic space This may be easily made responsible for the length phenomena in Icelandic compounds The presence of an empty CV assures that the FEN of the first member is ungoverned, hence it may licence the preceding nucleus

# 21 Government Phonology Round Table 9, Budapest, 20 April 2013

# 22 Government Phonology Round Table 9, Budapest, 20 April 2013

Advantages:

• the role of the morpheme boundary is recognised
• it materialises itself as a truly phonological object

Disadvantages:

• the motivation for translating the boundary into an empty CV is still

unclear

• it still appears to be dependent on melody

# 23 Government Phonology Round Table 9, Budapest, 20 April 2013

Comparison:

• Gussmann’s syllabification algorithm correctly recognizes the

importance of melody, but does not consider the role of boundaries

• Direct Interface: provides a way to represent the morpheme

boundary, but it is a mystery how melody can be a trigger for translation

A successful account of the phenomenon should be able to consider both melody and boundary information – it is clear that both play a role!

# 24 Government Phonology Round Table 9, Budapest, 20 April 2013

The point of departure: phonology cannot manipulate boundaries

• therefore Gussmann (1985) and Booij (1986) are unacceptable
• boundaries are provided by morphosyntax

rauð+leitur [ˈrœʏðlɛithʏr] ‘reddish’ and hvít+leitur [ˈkhvi:thlɛithʏr] ‘whitish’ must have the same kind of boundary there must be a phonological reason why the vowel is lengthened in one case but not the other the necessity of some kind of syllabification algorithm, which takes the boundary into account

# 25 Government Phonology Round Table 9, Budapest, 20 April 2013

The present proposal:

• couched within a modified model of Strict CV (introduced in

Fortuna 2013ab), which eliminates Proper Government and heavily enhances the role of licensing

• a blend of Strict CV with Cyran’s CSL + a few innovations
• clusters are established via Leftward and Rightward Interonset

Government

• both relations belong to computation, which invariably operates

right to left (=a kind of syllabification algorithm)

• both possible only when the consonant is Government-Licensed

by a nucleus

# 26 Government Phonology Round Table 9, Budapest, 20 April 2013

• Rightward Interonset Government – corresponds to branching
• nsets (steep TR clusters) and replaces Infrasegmental

Government

• Leftward Interonset Government: can be established in all other

clusters: RT, RR, TT, flat TR (in Strict CV with a nucleus silenced by PG)

• SGP and CSL three-way distinction (branching onsets vs. coda-
• nsets clusters vs. bogus clusters) eliminated – in terms of

phonological behaviour there are only two types of clusters, branching onsets vs. other

# 27 Government Phonology Round Table 9, Budapest, 20 April 2013

• all nuclei license both preceding onsets and preceding nuclei
• nuclei enclosed within LIO lose their licensing abilities
• hence, nuclei before LIO domains (=in closed syllables) end up

unlicensed

• this is responsible for closed vs. open syllable effects
• the presence of licensing is assumed to be a prerequisite for

vowel lengthening

# 28 Government Phonology Round Table 9, Budapest, 20 April 2013

• Direct Interface needs to be adjusted to the Icelandic data
• the output of translation is not empty syllabic space!
• Translator’s Office manipulates Final Empty Nuclei; it may choose

to flag a FEN as a ‘true FEN’

• if it doesn’t, the concatenation is synthetic (‘weak boundary’):

– both types of IO can be established across the cluster arising on a boundary – the cluster behaves exactly like a morpheme-internal cluster

• if it does, the concatenation is analytic (‘strong boundary’):

– the FEN after the first morpheme becomes a true ‘FEN’ and may block IO (To what extent? It’s language-specific!)

# 29 Government Phonology Round Table 9, Budapest, 20 April 2013

• LIO per definitionem can be established in all clusters which are

not the domain of RIO (regardless of sonority slopes and other factors), but there are language-specific deviations

Icelandic compounds:

• all of the examples are concatenated analytically
• so: FEN is a ‘true FEN’
• RIO is impossible across a true FEN
• LIO is possible across a true FEN only if the governee is not too

complex

# 30 Government Phonology Round Table 9, Budapest, 20 April 2013

[ð] is easily governable within LIO both across a ‘normal’ empty nucleus and a FEN It is of small complexity, so even FEN cannot save it Hence, FEN is governed and cannot license the preceding nucleus

# 31 Government Phonology Round Table 9, Budapest, 20 April 2013

but… [th] guarded by a FEN is too difficult a target for the following consonant to establish LIO across therefore, FEN is ungoverned and may license the preceding nucleus [th] is ungoverned itself and may be postaspirated

# 32 Government Phonology Round Table 9, Budapest, 20 April 2013

However, morpheme-internally LIO can always be established, no matter what consonant is the governee fortis plosives always undergo some kind of lenition when governed:

• preaspiration: /thth/ > [ht]
• spirantisation: /phth/ > [ft]
• deaspiration ~ spirantisation: /phs/ > [ps] ~ [fs]

the second consonant enforces LIO, therefore it leads to melodic depletion of its target

# 33 Government Phonology Round Table 9, Budapest, 20 April 2013

also /s/ (containing {H}) is governable morpheme- internally, but not across a FEN the patterning of /s/ with fortis plosives poses an important question: is it the matter of complexity calculus, or just the presence of {H}? both statements seem to work

• ther fricatives pattern with sonorants; probably they don’t

contain {H} /s/ is problematic in very many ways the question will be left open

# 34 Government Phonology Round Table 9, Budapest, 20 April 2013

Conclusion The paper provided an alternative way of handling post- lexical syllabification in Icelandic: a process which appears to violate modularity It was demonstrated that it is necessary to take into account communication between melodies: an argument against anti-melodism An alternative application of Direct Interface was proposed: if the Translator’s Office can upgrade the FENs and there is LIO (instead of Proper Government), it’s possible to explain why Icelandic compounds appear to contain sometimes one, sometimes two domains

# 35 Government Phonology Round Table 9, Budapest, 20 April 2013

Addendum: arguments for the model The presented model of phonology openly takes issues with anti-melodism present in Strict CV argument for anti-melodism: there are languages in which there are no restrictions on possible consonant clusters (e.g. Moroccan Arabic – Scheer 2004: 459-467) this can be incorporated within the present model: Moroccan Arabic chooses not to place restrictions on consonants involved in LIO – anything can govern anything but other languages do choose it! the existence of more specific phonotactic restrictions is inexpressible in models with Proper Government

# 36 Government Phonology Round Table 9, Budapest, 20 April 2013

for instance:

• coronals at word edges in English and German: Herbst, Arzt
• super-heavy rhymes in English involve only coronal clusters
• vowel ~ zero alternations which function differently in different

melodic configurations of neighbouring consonants (Scandinavian languages – Fortuna 2013ab)

these phenomena become much easier to explain when the responsibility for establishing clusters is shifted to consonants themselves

# 37 Government Phonology Round Table 9, Budapest, 20 April 2013

parametric choices include:

• is there LIO or not?

– if YES: what can govern what?

• is there RIO or not? (the existence of RIO presupposes the

existence of LIO! – 2010) – if YES: what can govern what?

• what is the influence of LIO or RIO on the intervening nucleus?

– default: LIO deprives the nucleus of licensing abilities, RIO does not – there may be language-specific deviations, may involve specific melodic primes! – ideally: the whole phonotactics of any language should be definable with these means – Probably a more elaborated theory of melodic primes is necessary

# 38 Government Phonology Round Table 9, Budapest, 20 April 2013

Bibliography

Árnason, Kristján. 2011. The Phonology of Icelandic and Faroese. Oxford: Oxford University Press. Charette, Monik. 1990. Licence to govern. Phonology 7: 233-253. Booij, Geert. 1986. Icelandic vowel lengthening and prosodic phonology. In F. Beukema & A. Hulk (eds.), Linguistics in the Netherlands 1986, 9-18. Dordrecht: Foris. Cyran, Eugeniusz. 2010. Complexity Scales and Licensing in Phonology. Berlin: Mouton de Gruyter. Fortuna, Marcin. 2013a. Icelandic and Faroese tonic lengthening in CV phonology: towards an alternative account. Paper presented at the ConSOLE XXI, Potsdam, 9-11 January. Fortuna, Marcin. 2013b. Why Icelandic and Faroese tonic lengthening is not a

• lengthening. Paper presented at the Old World Conference in Phonology 10,

Istanbul, 16-19 January. Gussmann, Edmund. 1985. The Morphology of a Phonological Rule: Icelandic Vowel Length. In: Gussmann, Edmund (ed.), Phono-Morphology. Lublin: RW KUL, 75-94. Gussmann, Edmund. 2002. Phonology: Analysis and Theory. Cambridge: CUP.

# 39 Government Phonology Round Table 9, Budapest, 20 April 2013

Gussmann, Edmund. 2006a. Icelandic vowel length and governing relations in

• phonology. Lingua Posnaniensis XLVIII, 21-41.

Gussmann, Edmund. 2006b. Icelandic and Universal Phonology. Greifswald: Ernst-Moritz-Arndt Universität Greifswald. Kiparsky, Paul. 1985. On the lexical phonology of Icelandic. In: Elert, C., I. Johansson, E. Strangert (eds.), Nordic Prosody III. Stockholm: University of Umeå. Oresnik, Janez. 1971. On the phonological boundary between constituents of Modern Icelandic compound words. Linguistica 11: 51-59. Scheer, Tobias. 2004. A Lateral Theory of Phonology. Volume I: What Is CVCV, and Why Should It Be? Berlin: Mouton de Gruyter. Scheer, Tobias. 2012a. Direct Interface and One-Channel Translation. A non- diacritic theory of morphosyntax-phonology interface. Berlin: Mouton de Gruyter. Scheer, Tobias. 2012b. Melody-free syntax and two Phonologies. Paper presented at the 10th Colloque du Réseau Phonologique Français (RFP), Paris, 25-27 June. Scheer, Tobias. 2013. Melody-free syntax. Paper presented at the Workshop on complexity at UQAM, Montreal, 8 February.

# 40 Government Phonology Round Table 9, Budapest, 20 April 2013

Takk fyrir! Thank you!