Metrical structure and Stratal Phonology provide a complete account of Danish stød Yonatan Goldshtein Pavel Iosad 29mfm, 25th May 2022 1 

Road map • A brief introduction to stød • Two key generalizations • Foot structure and stød • Domain structure and stratification • Analysis 2 

What is stød? 

The realization of stød piber ‘pipe-PL’ piˀber ‘squeak-PRS’ Sound files credit: Andrea Brink Siem 3 

Phonotactics • ‘Stød basis’ Stød can only occur on a stressed syllable with a heavy sonorous rhyme Syllable type No stød Stød CV nu ‘now’ * CV + obstruent kat ‘cat’ * CVV tale ‘speech’ råˀ ‘raw’ CV + sonorant kul ‘coal’ halˀ ‘hall’ CVV + sonorant team ‘team’ biˀl ‘car’ 4 

Distribution: anything goes? Singular Plural Singular definite Gloss biˀl biˀler biˀlen ‘car’ han hanˀner hanˀnen ‘male’ sumˀ summer sumˀmen ‘sum’ ven venner venˀnen ‘friend’ 5 

Key generalizations 

Basbøll (2005) et passim • The Non-Stød Model (Basbøll 2003; Basbøll 2005; Basbøll 2008; Grønnum & Basbøll 2001) 1. Stød is assigned to a stressed bimoraic syllable by default 2. Stød assignment can be blocked 2.3 Phonologically: lexical extrametricality 2.4 Morphologically: the Graded Productivity Model 6 

Our key generalization By default, stød is assigned to a stressed syllable at the word level, unless that syllable heads a disyllabic domain in the input to the word level • Cf. Itô & Mester (2015): stød is blocked when a (H ́ L) foot is coerced 7 

Domain structure and stød • Stød is phonotactically impossible in syllables without a heavy sonorous rhyme • Lexical extrametricality is a special case of this • Main stress feet are preferentially monosyllabic (H ́ ), except that stems containing a single foot at the right edge show (H ́ L) parsing 8 

Our basic assumptions • Stratal Phonology (Kiparsky 2000; Bermúdez-Otero 2012; Bermúdez-Otero 2018) • Base-Driven Stratification (Giegerich 1999) • √root + 𝑆𝐿 affix = stem-level phonology • [√root] 𝑆𝐿 + 𝑊𝐿 affix = word-level phonology • √root + 𝑊𝐿 affix = stem-level phonology 9 

Stød and stratification • Stød is assigned at the word level 10 

Stød and stratification • Stød is assigned at the word level • Stød is blocked if a disyllabic domain was constructed at the stem level 10 

Stød and stratification • Stød is assigned at the word level • Stød is blocked if a disyllabic domain was constructed at the stem level • Therefore: inner-attachment affixation blocks stød 10 

Analysis: monosyllabic nouns 

Word-level suffix attachment • No extrametricality: bil ‘car’ Singular Plural 𝑆𝐿 (bil) (bil) Stød? yes yes 𝑊𝐿 biˀl biˀler • Lexical extrametricality: han ‘male’ Singular Plural 𝑆𝐿 (ha) (ha) Stød? no: extrametricality yes 𝑊𝐿 han hanˀner 11 

Stem-level suffix attachment • Plural -e, not productive: hus ‘house’ Singular Plural 𝑆𝐿 (hus) (huse) Stød? yes no: (σσ) input 𝑊𝐿 huˀs huse • Plural -er, unproductive behaviour for this suffix: sum ‘sum’ Singular Plural 𝑆𝐿 (sum) (summer) Stød? yes no: (σσ) input 𝑊𝐿 sumˀ summer 12 

Stem-level suffix attachment continued • Now with extrametricality: ven ‘friend’ Singular Plural 𝑆𝐿 (ve) (venner) Stød? no: extrametricality no: (σσ) input 𝑊𝐿 ven venner 13 

Monosyllabic nouns with epenthesis • Extrametricality is irrelevant: stressed syllable coda is never word-final • Epenthesis/syllabification is word-level • Word-level attachment: bibel ‘bible’ Singular Plural 𝑆𝐿 (bibl) (bibl) Stød? yes yes 𝑊𝐿 biˀbel biˀbler • Stem-level attachment: finger ‘finger’ Singular Plural 𝑆𝐿 (fingr) (fingre) Stød? yes no: (σσ) input 𝑊𝐿 fingˀer fingre 14 

Exceptions • Two groups of nouns lack stød on a non-final sonorant mora, where it cannot be due to extrametricality • Certain historical clusters: mark ‘ground’, dirk ‘lock pick’ • Borrowings: team (contrast liˀm ‘glue’) • Also no stød in the plural • These have to be stored as word-level exceptions, e.g. via analytic listing (Bermúdez-Otero 2012) 15 

Disyllabic nouns 

Vowel-zero alternations • Not all nouns ending in a consonant + sonorant behave like bibel or finger • We analyse the following types as disyllabic with irregular syncope • Syncope is stem-level, explaining why it has exceptions (Kaisse & McMahon 2011) • Cf. Morrison (2019) on a similar pattern in Scottish Gaelic 16 

Word-level attachment • No syncope: helgen ‘saint’ • This is the regular pattern for disyllables Singular Plural 𝑆𝐿 (helgen) (helgen) Stød? no: (σσ) input no: (σσ) input 𝑊𝐿 helgen helgener • Syncope will not apply prior to affixation, so there is no counterpart pattern with syncope 17 

Stem-level attachment, syncope • Unproductive -e plural: himmel ‘sky’ Singular Plural 𝑆𝐿 (himmel) (himle) Stød? no: (σσ) input no: (σσ) input 𝑊𝐿 himmel himle • Unproductive stem-level -er plural: vabel ‘blister’ Singular Plural 𝑆𝐿 (vabel) (vabler) Stød? no: (σσ) input no: (σσ) input 𝑊𝐿 vabel vabler 18 

Stem-level attachment, no syncope • With -er, the predicted pattern is vabel ∼ vabeler • Indistinguishable from the word-level attachment pattern • Often attested as a variant for this group of nouns • With -e, the predicted pattern himmel ∼ himmele is unattested: gap? • The suffix -e is generally rare • Preference for -e to attach to surface monosyllabic bases 19 

Lexical monosyllabic feet • Underlying disyllables with lexical monosyllabic (stød-enabling) feet • No syncope: hummer ‘lobster’ Singular Plural 𝑆𝐿 (hum)mer (hum)mer Stød? yes yes 𝑊𝐿 humˀmer humˀmere • With syncope, the predicted pattern is humˀmer ∼ humˀre: indistinguishable from biˀbel ∼ biˀbler 20 

Other morphology • Unlike the plural, the definite clitic(s) (almost) always induce stød • Consistent outer attachment, as predicted morphosyntactically • Derivation • Inner-attachment, unproductive suffixes: lexical stød (cf. helgen vs. humˀmer) • Outer-attachment, productive suffixes: maintenance of stød, opacity (syˀ ‘sew’ ∼ syˀer ‘one who sews’) • Semi-productive suffixes: dual attachment possibilities (syˀning ‘sewing’ but rygning ‘smoking’) 21 

Compounds 

Regular patterns • Items with final stress regularly ‘lose’ stød when they are the first member in a compound • Monosyllables lose stød when non-final in a compound • huˀs ‘house’ ∼ husbåˀd ‘houseboat’ (N-N) • kriˀg ‘war’ ∼ krigsflyˀ ‘war plane’ (N-s-N) • rødˀ ‘red’ ∼ rødkåˀl ‘red cabbage’ (Adj-N) • finˀger ‘finger’ ∼ fingerringˀ ‘finger ring’ (N-N with epenthesis) • Stem-final stressed vowels shorten • industˈriˀ ‘industry’ ∼ industribyˀ ‘industrial town’ • Stød on a non-final syllable is preserved • raˀdio ‘radio’ ∼ raˀdiotårˀn ‘radio tower’ • Stød on a final syllable is preserved in longer roots • passageˀr ‘passenger’ ∼ passageˀrtogˀ ‘passenger train’ 22 

Exceptional patterns • In some compounds, exceptional stem-level constructs are stored nonanalytically, with maintained stød • landˀ ‘land’ ∼ landmandˀ ‘peasant’ (regular) ∼ landˀsmandˀ ‘compatriot’ (exception) • rødˀ ‘red’ ∼ rødkåˀl ‘red cabbage’ (regular) ∼ rødˀgrødˀ ‘berry porridge’ (exception) 23 

Our generalization, now in compounds Stressed syllables in compounds have stød unless they are fol- lowed by another foot at the stem level 24 

Stød loss: monosyllables • husbådˀ ‘houseboat’: stød not assigned to hus as it is followed by another foot at the stem level [(hus) Ft (båd) Ft ] 𝑆𝐿 25 

Stød ‘preservation’ • Items with antepenultimate stress (raˀdio) regularly get stød because (H ́ L) footing is disallowed • Exceptional items (rødˀgrødˀ, landˀsmandˀ, åˀbredˀ ‘river bank’) have stems with nonanalytically stored stød in the first element • Polysyllabic stems must project their own PWd: final monosyllabic foot gets stød [(passa(geˀr) Ft ) PWd ((toˀg) Ft ) PWd ] 𝑆𝐿 • These patterns show limited productivity, as expected 26 

Interim summary: nouns • The essential generalization is that stød is assigned unless the syllable is non-final in a stem-level domain • The morphophonology of stød lines up with morphological patterns: • Outer attachment: productivity, phonological opacity • Inner attachment: lack of productivity, phonological transparency • No recourse to bespoke domain structure (Basbøll 2005), but some role for prosodic optimization (Itô & Mester 2015) 27 

Verbs 

Verbs: summary of morphology Form ‘paint’ ‘talk’ ‘drive’ ‘see’ INF male tale drive seˀ PRES maˀler taˀler driˀver seˀr PAST malede talte drevˀ såˀ PTCP malet talˀt drevet seˀt IMP maˀl taˀl drivˀ seˀ 28 

Unprefixed verbs: analysis Form Level ‘paint’ ‘talk’ ‘drive’ ‘see’ INF 𝑆𝐿 (mal-e) (tal-e) (driv-e) (se-e) → (se) 𝑊𝐿 male tale drive seˀ PRES 𝑆𝐿 (mal) (tal) (driv) (se) 𝑊𝐿 maˀler taˀler driˀver seˀr PAST 𝑆𝐿 (mal-e) (tal-te) (drev) (så) 𝑊𝐿 malede talte drevˀ såˀ PTCP 𝑆𝐿 (mal-e) (talt) (drev-et) (se-et) → (set) 𝑊𝐿 malet talˀt drevet seˀt IMP 𝑆𝐿 (mal) (tal) (driv) (se) 𝑊𝐿 maˀl taˀl drivˀ seˀ 29 

Prefixed verbs • Prefixed verbs have stød on the root even where the unprefixed one does not • tal-e ‘speak-INF’ ∼ udtaˀl-e ‘pronounce-INF’ ∼ betaˀl-e ‘pay-INF’ • Contrast the behaviour of nouns under inner attachment • hus-e ‘house-PL’ ∼ udhus-e ‘outhouse-PL’ • udtale ‘pronunciation’ • Both stem-level constructs! • Bracketing paradox? No, prosody • [[ud-talˀ] 𝑆𝐿 -e] 𝑆𝐿 : final in inner domain, regular cyclicity ⇒ stød • [(ud) PWd -((hus-e) Ft ) PWd ] 𝑆𝐿 : prosodic requirements of the prefix trigger construction of PWd and (H ́ L) foot per the usual generalization ⇒ no stød 30 

Summary and prospects 

Overall conclusion • The behaviour of stød emerges from relatively simple generalizations • Mono- vs. disyllabic domains • Familiar moraic phonology: stød basis, extrametricality • Stratal Phonology with Base-Driven Stratification • Our analysis captures the phonological consequences of attachment asymmetries with no extra stipulations 31 

Why domain size? • Why would non-final stressed syllables reject stød? • One possibility, following Köhnlein (2016): head vs. non-head morae • In a monosyllabic (H ́ ) foot, the stressed syllable is 𝜇+𝜇− • In a disyllabic (H ́ L) foot, the stressed syllable is 𝜇+𝜇+ • Laryngealization can only be assigned to non-head morae because of positional faithfulness (Iosad 2016) 32 

Wider context • Stratal Phonology accounts for morphology-phonology interactions in Danish stød • ‘Accentual’ distinction between two types of syllables accounted for via domain (foot?) structure • Cf. ‘metrical’ approach to tonal accents in Germanic (Hermans 2009; Morén-Duolljá 2013; Köhnlein 2016), Scottish Gaelic (Morrison 2019), Finnic (Odden 1997) 33 

Mange tak! [email protected] [email protected] 33