University of Southern Denmark

Danish stød

towards simpler structural principles?Grønnum, Nina; Basbøll, Hans

Published in:Understanding Prosody – The Role of Context, Function, and Communication

DOI:10.1515/9783110301465.27

Publication date:2012

Document version:Final published version

Citation for pulished version (APA):Grønnum, N., & Basbøll, H. (2012). Danish stød: towards simpler structural principles? In O. Niebuhr (Ed.),Understanding Prosody – The Role of Context, Function, and Communication (pp. 27 - 46). De Gruyter.Language, Context and Cognition Vol. 13 https://doi.org/10.1515/9783110301465.27

Go to publication entry in University of Southern Denmark's Research Portal

Terms of useThis work is brought to you by the University of Southern Denmark.Unless otherwise specified it has been shared according to the terms for self-archiving.If no other license is stated, these terms apply:

• You may download this work for personal use only. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying this open access versionIf you believe that this document breaches copyright please contact us providing details and we will investigate your claim.Please direct all enquiries to puresupport@bib.sdu.dk

Download date: 24. Jul. 2021

Language, Context, and Cognition

Edited by

Anita Steube

Editorial Board: Kai Alter· Ulrike Demske· Ewald Lang

Rosemarie Liihr · Thomas Pechmann · Richard Wiese

Volume 13

De Gruvter

Understanding Prosody

The Role of Context, Function and Communication

Edited by

Oliver Niebuhr

De Gruyter

FSC

MIX

Paplor aus vora.ntwor­tungsvollon QuoUon

FSC- C016439

ISBN 978-3-11-030125-0

e-ISBN 978-3-11-030146-6

ISSN 1866-8313

Library of Congress Cataloging-in-Publication Data

A CIP catalog record for this book has been applied for at the Library of Congress.

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet

at http://dnb.dnb.de.

© 2012 Walter de Gruyter GrnbH, Berlin/Boston

Printing: Hubert & Co. GmbH & Co. KG, Gottingen oo Printed on acid-free paper

Printed in Germany

www.degruyter.com

Contents

Preface ................................................................................................ VII

Yi Xu & Fang Liu Intrinsic coherence of prosodic and segmental aspects of speech................................................................................................. 1

Nina Grannum & Hans Basbo/1 Danish st0d - Towards simpler structural principles? ............... 27

Ilse Lehiste Prosodic allomorphs in the Estonian declension system.......... 47

Han.!}or;g Mixdorf! The application of the Fujisaki model in quantitative prosody research ............................................................................................. .

Amalia Arvaniti Rhythm classes and speech perception ........................................ .

Jonathan Barnes, A!tjna Brugos, Stefanie Shattuck-Hefnage! & Nanette Veilleux On the nature of perceptual differences between accentual peaks and plateaux .......................................................................... .

Susanne Scho'f:v Costa Bruce, My Segerup, Jonas Beskow, Joakim Gustafson & Bjo'rn Granstrom Regional varieties of Swedish: Models and synthesis ................ .

GilbertAmbrazaitis,Johan Frid & Costa Bruce Revisiting Southern and Central Swedish intonation from a comparative and functional perspective ...................................... .

55

75

93

119

135

Nina Grnnnum (Copenhagen) & Hans Basb0ll (Odense)

Danish st0d -Towards simpler structural principles?

1 Introduction

St0d is a well established phenomenon in Danish. The first scientific account of it is to be found in Jens Pedersen H0ysgaard's (1698-1773) Concordia res parvce crescunt, e//er Anden Prove cif Dansk Orthographie (H0ysgaard 17 43) and Accentuered og Raisonnered Grammatica (H0ysgaard 17 4 7). However, an indirect reference to st0d occurred already 200 years earlier in an inflammatory speech against the Danes, held in 1510 by the Swedish bishop Hemming Gadh: Vehemens contra Danos oratio. The source may be historically inaccurate, but the speech is contained in Johannes Magnus' (1488-1544) Historia de omnibus Gothorum Sueonumque regibus. Magnus' opus is most likely composed before 1526 - when he left for Rome never to return - but was only published in 1554 (by Olaus Magnus). The pertinent part ofGadh's speech reads thus (Magnus 1554, 2nd ed. 1617, p. 875):

''Nee ut creteri homines loqui dignantur, immo more tussientium, aut verba in medio gutture formantium, ita de indust<r>ia proloquuntur, ut superius labium in sinistrum, inferiusque in dextrum latus distorquentes, ex singulari oris deformitate, singularem gloriam sese as?equi posse existiment."

A Swedish translation from some time later in the century (reprinted in Soderberg 1908) reads:

"Der till medh: sa wrerdas de icke heller att talla som annat fold" uthan tryckia ordhen fram lika som the willia hosta, och synas endeles meclh flitt forwendhe ordhen i strupan, for ren de komma fram, sammaledes wanskapa the munnen, d:l. the talla, wridhan och wrengan, s:l. att the draga then offurra leppen till then wenstta sidon och den nedra till then hogra sidon, menandes dett wara sig en besynnerlighe prydning och wellst:l.ndh."

In approximate present-day English that would be:

Not only do they not stoop to speak like other people, but press the words out as if thry will cough, and appear even deliberatefy to twist the words in the throat befare thry come out, likewise they misshape the mouth when they speak, twisting and

28 Nina Gr0nnum & Hans Basb0!1

turning it, pulling the upper lip to the left and the lower lip to the right, deeming that to be particularly becoming and proper.

(Italics single out the passage which we take to be an indirect reference to the st0d).

St0d is absent in a number of Danish dialects and regional languages, namely roughly south of a boundary drawn from R0m0 in the west via Faborg and Nyborg on Funen across the southernmost tip of Zealand to Prrest0 in the east, cf. Ejskjrer (1990). When and where st0d occurs it does not necessarily have the same distribution as in standard Copenhagen Danish. We are dealing here exclusively with the standard Copenhagen variety.

2 The phonetics of st0d

St0d is a kind of creaky voice, that is non-modal voice with aperiodic vibrations and irregular amplitude. It is therefore found exclusively invoiced sounds, and it only occurs under conditions of more than zero stress. Most noteworthy, it requires a certain minimum amount of voiced material in the syllable rhyme in order to be present:

(a) either a long vowel, as in [phe:1n] peen 'nice'

(b) or a short vowel+ a sonorant consonant, as in [phen1] pen 'pen.'

(St0d is conventionally marked after the long vowel symbol and after the first post-vocalic sonorant following short vowels, respectively. For lack of a more appropriate notation we use a superscript [ 1 ] .)

Syllables which fulfill these segmental and prosodic requirements are heavy in Danish phonology. That is how they were characterized in Basb0ll (1988), where the notion weight unit in Hyman's sense (1985) was employed. In later works they are called bi-moraic (see, for example, Basb0ll 2005, 2008). The idea of st0d as a mora-counting device can also be found in the work of some Prague phonologists (f rubezkoy 1935: section 34, and Martinet 1937: 100-102); see also Liberman (1982). Fischer­J0rgensen's (1987, 1989a, 1989b) phonetic description of vowels with st0d as having two distinct phases, a preparatory non-st0d phase succeeded by a st0d phase proper, is in accordance with such a moraic analysis.

Basb0ll's (1998) mora-analysis - and Fischer-J0rgensen's previous investigations - posed a number of questions about consonant and vowel duration and about the perception of st0d. In the early 2000s we addressed some of these questions and found:

Danish st0d- Towards simpler structw:a.1 principles? 29

(1) Vowels with st0d are as long acoustically and perceptually as longvowels without st0d. Accordingly, st0d vowels could be bi-moraic(Grnnnum and Basb0ll 2002a, 2002b).

(2) Consonants with st0d are not generally longer acoustically than

consonants without st0d across all positions. If consonants withst0d are moraic and consonants without st0d are not moraic andif morae in Danish are to have durational correlates U.: theconsonants - as they do in typical mora-counting languages, this isan obstacle for the analysis (Grnnnum and Basb0112001a, 2001b).

(3) Listeners generally perceive the st0d onset in long vowels tocoincide with vowel onset. In other words, there is no perceptualbi-partition with st0d confined to the second part of long vowels(Grnnnum and Basb0ll 2003a, 2003b).

An unambiguously two-phased st0d would imply that the creah.-y voiceshould be contained in the second half of long vowels or within the sonorant consonant after short vowels. Such a well-defined acoustic align­ment may occur, as in the example in Figure 1 left, but mostly it does not, cf. the example in Figure 2. St0d may even be acoustically quite elusive, as in Figure 3 left. In other words:

( 4) The exact acoustic properties, the timing, and the segmentaldomain of !he st0d are highly variable.

One interesting thing about this considerable acoustic variability is that it does not seem to affect the perception of st0d: St0d is as clearly audible and identifiable in the word in Figure 3 left as in the words in Figure 1 left and Figure 2. For a more comprehensive account, see the explanations and examples in Grnnnum and Basb01l (2007).

30

5000

-;;-� 4000

g 3000

2000 e

Nina Grnnnum & Hans Basb0ll

F;"T

u.

,oao :::'J'iiib:f:��:�:·.::::�: ....

.......... ..0 -i I

···············1·;··..,·················'ti'·····•··•······

.... ········. . .. . .. . . .. ·····1·

o.s Time (s) 1.0 1.4

Figure 1: Spectrogram and waveform of a word, Ballet The Sound' (a proper name), with st0d in

the sonorant consonant (left) and ballet 'the belt' without st0d (right). The waveform of the

consonants has been expanded for easier viewing in the lower part of the figure. Female speaker.

�ooo .....

N ,!OQQ:.

� � 3000 <: :, er

LL

2000 ....

1000.

.. P. - 1-···· ······i··· ····1···· ···I···

0.1 0.3

. . . I • • • • • • • . . . • .

0.4

Figure 2: Spectrogram and waveform of a word, /aner 'leans,' with st0d (111 the long vowel) spilling

over through the initial consonant into the vowel of the succeeding syllable. Female speaker.

� <: .,

LL

Danish swd - Towards simpler structural principles?

5000 -·

·�

4000 -

I 3000 -·

l I/ 1 ··2000 -· ,, 1000 -·

o -1 .....

0

1�111,.I ;l!i ',\'4

Time(s) 1.0 1.3

[ 1sgre:9e]

II!�

�1�iif1V1,-

31

Figure 3: Spectrogram and waveform of a word, skaber 'creates', with st0d (left) and skaber

'creator' without st0d (right). The waveform of the vowels has been expanded for easier viewing in the lowe.r part. Male speaker.

2.1 St0d as a ballistic gesture

Danish st0d is very reminiscent perceptually of the glottalization found in German as described by Klaus Kohler (Kohler 1994, 2001). Its function is different, of course, Danish st0d being phonologically and morphologically conditioned and distinctive, cf. sections 3 and 4 below. Presumably, a.t least certain aspects of its production would also keep it apart: Fischer­] 0rgensen (1987, 1989a, 1989b) contains an account of various phonation types, particularly creah.-y voice. She concludes that many features are iden­tical in creah.-y voice and st0d, but that st0d is not simply creaky voice. We agree entirely, and we speculated about the kind of articulatory mechanism which could be made responsible for the rather astounding acoustic - if not perceptual - variability, cf. above. This is what we came up with in Gr0nnum and Basb0ll (2007) as a characterization of the st0d in articula­tory terms:

32 Nina Gr0nnum & Hans Basb0ll

(1) The laryngeal activity is a ballistic gesture which - minimally -makes for a slightly compressed voice quality, at one end of acontinuum, and - maximally - creates a distinctly creaky voice atthe other. Under emphasis it may become a complete glottalclosure.

(2) It is a property of the (sonorant part of the) syllable rhyme.(3) It is aligned with the onset of the rhyme.(4) It is variable with respect to strength and to temporal extension.The proposed ballistic gesture is to be understood as the low-pass

filtered muscular response to a transient neural command. The neural command is presumably timed to coincide with the onset of the syllable rhyme. The impulse may be stronger or weak.er, resulting in more or lessi irregular vocal fold vibration of shorter or longer duration, but once the command is executed, the speak.er can no longer control the way the vocal folds respond to the excitation, just as one can no longer control the trajectory of a tennis ball once the ball has bounced off the racket. This proposal is consistent with the fact that speakers cannot choose to increase the duration of the st0d ad libitum the way one may choose to lengthen creah.-y voice at the end of an utterance. It is consistent as well with the way we have seen the st0d to behave acoustically, cf. the more explicit (Fig. 1 left and Fig. 2) or less explicit (Fig. 3 left) non-modal vocal fold vibration; the variable timing of the onset of actual creah.-y voice in the waveform and the spectrogram (Fig. 2); and the variable total duration which often makes it continue well into the following syllable (Fig. 2). Furthermore, our proposal is consistent with EMG-data: the onset and offset of the higher vocalis muscle activity in st0d relative to modal voice are executed as a smooth, gradual rise and fall (Fischer-J0rgensen 1987, 1989a, 1989b). It would be curious indeed if the actual mechanical change in vocal fold vibration mode were not also gradual. At present we have no indication that the variability in strength and timing is not random. However, investiga­tions of st0d in a corpus of non-scripted speech (Gr0nnum 2009) may reveal individual differences among speakers and variation as a function of speech rate, or - more likely - degree of prominence on the syllable:

Given the results of our investigations and our hypothesis about the underlying physiological mechanism, Basb0ll's (1988) proposal of 'st0d as a signal for the second mora of syllables' is not an acoustic nor an immediate cognitive reality. His mora analysis has undergone significant changes in

recent years: extra-prosodicity (that is: position outside a given prosodic frame) has become a central concept, and with it he has been able to make a number of important predictions about st0d and lexical patterning (Basb0ll 2003, 2005: 400-14, 2008), as evidenced below.

Danish seed - Towards simpler structural principles?

3 The :function of st0d

St0d is indisputably distinctive on the surface:

['lc:7 SB] kzser 'reads' vs. ['le:sB] kzser 'reader' ['vre:71I).J hvalen 'the whale' vs. ['vre:lI).] valen 'numb'['hu:7s9] huset 'the house' vs. ['hu:s9] huset 'housed'['t'cem 7B] temmer 'timber' vs. ['t'cemB] tommer 'empties'['hm7B] hamder 'hands' vs. ['hmB] hamder 'happens' ['s<Jex/B] stanger 'rods' vs. ['s<JEIJB] stanger 'locks up' ['gdB] galder 'is valid' vs. ['gelu] gal/er 'gills'

33

(Note that the vowel [B] in the examples above is the manifestation of I ar I, as in, for example, ['lc::sB] laser 'reader'; and I a I assimilates to a neighbouring vowel or sonorant consonant as in, for example, ['lpru:u] bruge 'use,' ['vre:1h;1] hvalen 'the whale.' Likewise Ira I and I rar I also coalesce in [B], as in, for example [ 1

0IJiro] angre, angrer '(to) regret,' '(she) regrets.')

Although st0d distinguishes meanings, it is, as we shall see below, to a very large extent predictable from syllabic and morphological structure.

The st0d/ non-st0d distinction roughly parallels the Accent I/ Accent II distinction in standard Swedish and Norwegian. Words with st0d in

Danish generally correspond to words with Accent I in Swedish and Norwegian, and words without st0d correspond to words with Accent II. There are some nota�le differences, however.

(1) The st0d/non-st0d distinction is a difference in creah.-y versusmodal voice quality; the Swedish and Norwegian word accentdistinction is purely tonal.

(2) There are segmental restrictions in st0d occurrence, cf. section 2above; no such restrictions apply to the word accents.

(3) St0d is associated with syllables, not words; the opposite is true ofword accents. It takes at least two syllables in Swedish orNorwegian for Accent II to occur; furthermore, in compositewords Accent II will stretch out and cover the whole word.

(4) Danish monosyllables may have st0d or not; monosyllables in

Swedish and Norwegian always carry Accent I.(5) Swedish and Norwegian stressed syllables are always heavy, either

because the vowel is long or because a short vowel is followed bya long postvocalic consonant. In Danish, the heavy versus lightdistinction in stressed syllables with short vowels depends on thenature of the coda consonant: only a sonorant consonant will

34 Nina Gr0nnum & Hans Basb01l

make the syllable heavy, as, for instance, 1n [sg.:rrnm\'-J stramt

'tight' vs. [lags] laks 'salmon'. Together these differences make any direct comparison of tonal word

accent in Swedish and Norwegian with Danish st0d phonetically, phonologically and morphologically opaque. In addition, to our knowledge, no trends have been reported for Swedish or Norwegian to match our observations and speculation below about the new trends for st0d dis­tribution in section 6 below.

4 Phonology and morphology of st0d

Initially, we should point out that the principles governing st0d are productive. This is reflected, inter alia, in the way most Danes pronounce German and Austrian composers, for example:

['mo:so:1 g.] Mozart rather than ['mo:t' curt] ['hm7g.n Handel rather than ['hmdn [lnm:1ms] Brahms rather than [brm:ms].

Latin and Greek words likewise obey the st0d principles of the native Danish vocabulary, whereas English and French loans generally obey the principles oflexical non-st0d (cf. Basb0ll 2008: 155-160). We consider only the native st0d principles here.

4.1 St0d and word structure

This section presents some instantiations of general principles stated in Basb0ll (2003, 2005). A general, operative principle of st0d is that:

(1) lexically specified properties remain constant throughout.

That is, if a lexical item is marked with st0d, the st0d will appear in every inflected and derived form, and it is not subject to deletion under any circumstances (except due to stress reduction as, for example, in composition or in unit stress reduction, cf. Rischel 1983). Conversely, if a lexical item is marked for non-st�d, st0d does not turn up in any context. Such well-established lexicalized forms are not considered any further here, because they are not the result of productive processes, and they are immaterial to our presentation of st0d in unexpected contexts.

Danish st0d - Towards simpler structural principles? 35

4.1.1 St0d in non-infle,cted, non-derived words (lexical items)

As noted earlier, st0d only occurs in segmentally heavy syllables with more than zero stress.

[mu:1s] mus 'mouse' [phan? g.] pant 'lien' [vom7s] vams 'doublet'

['mu:sa] muse 'muse (n)' ['t'ang.a] tante 'aunt' ['gomsa] bamse 'teddy-bear.'

From the examples above we would conclude that monosyllabic lexical items have st0d, disyllabic lexical items do not. However, consider

[ga'lan7g.] galant 'chivalrous' [l>a'la.IJS:l] balance 'equilibrium' [ ela'fan7g.] elefant 'elephant'

The disyllable ga/ant has st0d; the tri-syllable balance is without st0d; and the tri-syllable elefant has st0d. The proper general principle now appears to be:

(2) the penultimate syllable of lexical items has no st0d.

There are exceptions to this Rrinciple, however, in some lexical itemsending in 1�11, [�n[, [�rl, cf. [1£IJ gn enke/ 'simple,' [1vo:'gn;t] vaben 'weapon,' ['il?g.-e] ilter 'short-tempered,' but not all, cf. [ 1£1Jn engel 'angel,' [ 10:gn;t] aben 'open,' [1al9-e] alter 'altar.' Most of the lexicalized forms with st0d derive historically from monosyllables. In accordance with (1) above, they do not undergo morphological st0d-altemations.

4.1.2 Inflection and derivation

Matters are more complex in inflected and derived words due to the differ­ent behaviour of suffixes, depending on their degree of productivity (Basb0ll 2003, 2005: 351-63):

(3) Suffixes are fully productive, semi-productive or non-productive.

Basb0ll's categorization of suffixes and its consequences for his account of st0d behaviour entailed a considerable simplification, compared to pre­vious descriptions, when he turned matters on their head and showed how the pertinent question no longer is which syllables have st0d, but rather:

36 Nina Grnnnum & Hans Basb0ll

( 4) When does a heavy syllable not have st0d?

The answer is bi-partite:

( 4a) It depends on the productivity of the suffix. (4b) Before semi-productive suffixes it also depends on the stem:

monosyllabic versus polysyllabic.

We will look at one inflectional and one derivational suffix from each of the three productivity groups for illustration. That will suffice to Jgive the reader an idea of the principles regulating st0d in inflected and derived forms.

4.1.2.1 Productive suffixes

An example of a productive inflectional suffix is the plural morpheme I �r I, cf.

[vre:11] hva/'whale (sg.)' [ ela'fan1cJ] elefant 'elephant (sg.)' ['t'an<Ja] tante 'aunt (sg.)' [fa'jOJJS:l]jqjance 'faience (sg.)'

['vre:711?] (pl.) [ela'fan7<j"B] (pl.) [1t'an<j1?] (pl.) [ fa1jOJJS'B] (pl.)

(Note that \;)\ + \ � \ contract to \;)I, thus, for instance, ltant;)I + l�rl > ltant;)rl, and I ;)fl > [u].) A productive derivational suffix is noun forming \eng nengl, cf.

[ sy:7] sy 'sew' ['sy:7neIJ] !Jlning '(the act of) sewing' ['vana] vande 'water (v)' ['vanel)] vanding 'watering (n)' (!al is dropped before the vowel-initial suffix.)

The lexical items with st0d, hva4 elefant, !J, retain their st0d, and the st0d­less !ante, fqjance, vande remain without st0d, because:

(5) Productive suffixes have no effect on st0d as such.

There are many ways to conceptualize this fact: one might be that the productive suffix is so loosely attached to the stem that stem and suffix do not interact.

Danish st.0d - Towards simpler structural principles? 37

Note, however, that an added suffix may become integrated in the stem and modify its moraic structure, to the effect that the word receives st0d. Thus, for example, han [han] 'male (n)' is without st0d because the [n] is (lexically specified as) extra-prosodic. In the plural, lhanl + l;)rl, thestem-final \n I cannot be extra-prosodic since it is no longer word-final(Basb0ll 2005: 388-93), and accordingly we get ['han2u] banner 'males' with

. st0d.

4.1.2.2 Non-productive suffixes

An example of a non-productive inflectional suffix is the noun plural morpheme \� \ , cf.

[hun7] hund'dog (sg.)' [hu:7s] hus 'house (sg.)'

vs. vs.

['huna] (pl.) ['hu:sa] (pl.)

A non-productive derivational suffix is the noun forming /m;)/, cf.

[1s0oma] sodme 'sweetness'

The ensemble behaves like a lexical item and principle (2) is operative: the penultimate syllable of lexical items has no st0d. This translates into yet a principle:

(6) A non-productive suffix is integrated in the stem.

The integration of stem and suffix may be conceived as the effect of a weak boundary between them. It stands to reason that a productive suffix has a more autonomous status cognitively, and is more easily separable from the stem, than a non-productive one.

4:1.2.3 Semi-productive suffixes

An example of a semi_-productive inflectional suffix is infinitive l�I, cf.

[('l;rrru:a >) 'g:iru:u] bruge 'use' [('fala >) 'fatnfalde 'fall'

vs. ['gmg:iru:1u] genbruge 'recycle' vs. ['nW"Bfafn overfalde 'attack'

A semi-productive derivational suffix is, for example, adjectival \i\, cf.

38 Nina Grnnnum & Hans Basb0ll

[mo:15] or [moo7] mod'courage' vs. ['mo:oi] modzg 'courageous' vs. ['owt?mo/oi] ovennodig 'fool-hardy'

It appears that

(7) Before semi-productive suffixes only monosyllabic stems have nost0d. i

But a st0d appears when the stem is expanded to the left. Note that the only two semi-productive inflectional suffixes are verbal, namely infinitive \::} \ and preterite \ t� \, whereas there are a dozen semi-productive deriva­tional suffixes.

The net result of principles (2) through (7) is that phonology and morphology together, by and large, predict the presence and absence of st0d. Conversely, st0d and its absence will act as a cue to morphological structure. Note especially that these principles are not sensitive to word class, but exclusively to word structure (and syllabic structure) and degree of productivity of the (inflectional or derivational) suffix.

5 Psycholinguistic issues

There are some less tangible aspects of the nature of st0d which should complete the picture of this fascinating phenomenon, even though we cannot back up all of them with empirical evidence:

St0d is apparently not an obstacle where children's acquisition of their mother tongue is concerned. That is, words with st0d enter into a child's lexicon no later than corresponding words without st0d; nor do st0d alternations (for example from singular to plural) delay children's acquisition of morphology, cf. Kjrerba::k and Basb0ll (2010: 15, 25).

In connection with the phonetic annotation of a fairly large non­scripted speech corpus, DanPASS (Grnnnum 2009), the transcribers often noted that - although the acoustic manifestation was highly variable, as demonstrated in section 2 above - st0d was very nearly always clearly identifiable, and - even more noticeable: it did not seem to be subject to any particular weakening, whether acoustic or perceptual, in less distinct and/ or more rapid passages. This is in sharp contradistinction to the manifestation of most segments, of course, which - while still being indisputably identifiable - display a large range of varieties. Thus, for instance, obstruents vary from the most clearly enunciated proto-typical stops and fricatives to the weakest possible approximants, and they may be deleted altogether. St0d is never likewise deleted (the morphological non­w;,1<1 nrinr.inle.s are. a c1ifferent. matter. cf. section 4.1.2 above).

Danish st0d - Towards simpler structural principles? 39

Slips-of-the-tongue involving st0d are extremely rare. In Nina Grnnnum's collection from radio transmissions of hundreds of slips, there are only two involving st0d. This is one of them: . . . Berlusconis kvaler og skandaler ... ' . . . Berlusconi's troubles and scandals ... ': kvaler was rendered with st0d, ['khvre:7le], as it should be, and so was, mistakenly, skandaler[sgan\lre:7le], supposedly in a carry-over from kvaler.) In this respect, st0d resembles stress which also only produces very rare slips-of-the-tongue.

In a pilot phonological experiment which, among other tasks, involved syllable reversal (Gr0nnum 1999), subjects would reverse the segments, but almost consistently leave stress and length in place. Thus, for example, ['mo:na] > ['nre:mo] (rather than *[na'mo:]) with adjustment of the short [a)-quality to the appropriate long [re:]-quality. Stress and length appear to be autosegmental, and the prosodic properties of words are presumably stored separately in the mental lexicon of speakers. This is in accordance with Stemberger's (1984) analysis of German and Swedish speech error data. It is also reminiscent of Hombert (1986) who showed that length (and likewise tones) remain in place in syllable reversal experiments. To these experimental findings we might add entirely anecdotal evidence: when a word will not come to mind immediately, we are nevertheless often able to recall its structure in terms of its number of syllables and the location of its stress - another kind of evidence that the word's prosodic frame is represented· cognitively separately from its segments.

Now, st0d is undoubtedly also a prosodic property. What may we conjecture about its representation in the lexicon? Regrettably, ma_terial for a syllable reversal experiment is not so easily construed as in the case of stress and length, because the two syllables to be reversed must both be heavy, and there are not many lexemes which fulfil that condition in Danish. The requirement would be met easily in compound words. We would then have to assume, however, that the compounds are separate, independent_ lexical entries, each with its integrated, separate prosodic frame. Such an assumption may well be valid if we are dealing with existing and common words in the lexicon but not, of course, if we were to construct nonsense words for the experiment. However, be that as it may, composita will not resolve the issue anyway: When, say, ['lanman1] landmand 'farmer' (literally: 'land man') becomes ['manlan7], which is our intuition, is that because st0d remains in place in the prosodic frame? Or is it because the new word is subjected to a well-established non­st0d principle: mono-syllabic first members lose their st0d ( cf. section 6 below), while land in the second position, after reversal, just retains the st0d it has in isolation?

40 Nina Gr0nnum & Hans Basb0Jl

If compounds will not work, we are left with lexemes ( or nonsense words which could be acceptable new lexemes). But that is not straight­forward either: The proto-typical Danish disyllabic lexeme has no st0d on any of its syllables ( cf. section 4.1.1 above) so there is no st0d whose position may or may not be affected by a syllable reversal. Less proto­typical disyllables can be found among proper names and place names which historically may be compounds but are hardly recognized as such to­day, like, for example, Anholt, Pano, Fausbo/4 Ingo(f ['anh.d<J], ['fre:110:1], ['fow2sg0l], ['eIJgAff]. Arrived at this point, we are forced to admit that our intuition about the outcome of a syllable reversal experiment is not a strong one. This experiment - or some other experiment which may tell us something about st0d and its cognitive representation - needs to be done. We shall have to leave the question hanging while we propose a working hypothesis: st0d in mutilated words will, to the extent possible, adhere to known principles for st0d and non-st0d in Danish words. If that is so, st0d is not exactly on a par with the other prosodic properties, stress and length, in Danish.

6 St0d in new and unexpected contexts

The principles for st0d assignment appear to be in the process of change, in the direction of simplification and generality, as indicated by data, mostly from the Danish Radio, Channel 1, collected by Nina Grnnnum over the past decade.

We should note first that compounding (as opposed to inflection and derivation) does not generally and in an unambiguously principled manner entail neither deletion nor addition of st0d. With two exceptions: (1) Loss of st0d may occur in a first member if it is monosyllabic and, in somewhat simplified terms, if this first member is familiar and well-established as such in the language. Thus, for example, so4 so/skin ['so:71], ['so:lsgen7

]

'sun,' 'sunshine,' but chef, chef/on ['<;;e:7f], ['<;;e:7flcen7] 'boss,' 'boss' salary'

(skin and Ion have st0d also in isolation). Addition of st0d is common in a final member of a compound if it is a polysyllabic verb or verbal derivative. Nor is this a new phenomenon. Hansen (1943) lists numerous examples, predominantly infinitives and verbal adjectives, for example faregogle

['fo:"BgCEj 1l�J 'pretend,' modstrcebende ['moosQ-I>re:7\;n;m�] 'reluctant;' gogle

'joke (v)' and strcebende 'striving' in isolation are without st0d. There is in this suspension of the non-st0d principle an analogy to principle (J) above (which as formulated there only accounts for non-compound words).

Danish st0d- Towards simpler structural principles? 41

However, st0d in this type of compound is not exceptionless. Hansen (1943) lists a number of instances - less common words in the vocabulary

according to him - where st0d vascillates; for example dodbringende

[1d00\):r,re1]<7li;m�] (literally:) 'deathbringing.' It is our distinct impressionthat st0d in this type of compound - where the final part is a verb or a verbal derivative - is becoming the rule rather than the exception in present-day Danish, cf., for example - from Nina Grnnnum's collection -hvalpelignende, Jesustroende [ 1val\)�li:1i;m� ], ['je:sustsK0:7:i:in�] (literally:)'puppyresembling,' 'Jesusbelieving,' which are definitely new words and -we suspect - actually created by the speaker on the fly. Likewise, among the different classes and subclasses of words in the data, those involving verbs or their derivatives contain by far the largest numbers of items.

Below we shall illustrate four cases where st0d is unambiguously 1.µ1expected in the standard language and which demonstrate the need for a revision of Basb0ll's (2005) model if it is to encompass a norm where such pronunciations have become the rule rather than the exception.

6.1 Simple nouns in the plural

['fo:mu:1-e]jommer 'fortunes' but [('fo:mu:� >) 'fo:mu:u].fo1711ue is without st0d

in the singular in the standard norm. [1Allll>o:1c>-e] omrader'areas' but ['AIIllfO:o�] omrade is without st0d in thesingular in the standard norm.

(However, although singular fa1711ue and omrade do not happen to occur with st0d in Nina Grnnnum's collection, st0d would no longer surprise us, given the uninflected lexical items that actually do occur with st0d, cf. below.)

This is surprising in light of principle (5), cf. section 4.1.2.1: Productive suffixes have no effect on the st0d, and plural j �r j ([ '8 ]) is productive. If one would argue that perhaps the boundary between this lexical item and the suffix has weakened, then we would be dealing with a penultimate syllable in a (pseudo-) lexical item, and it should have no st0d anyway.

6.2 Compound nouns in the plural

['vi:nnaw'n�] vinnavne 'wine names' but ['nawna] navne alone is without st0d; ['sy:yhu:'s�] sygeh11se 'sickhouses' (that is: hospitals) but ['hu:s�] huse alone is without st0d.

42 Nina Gr0nnum & Hans Basb0ll

Here is an attempt to account for these unexpected noun plurals: The stems - whether simple or compound - have two things in common:

they have endings they are polysyllabic.

Perhaps principle (7) - which concerns semi-productivi suffixes, cf. section 4.1.2.3 - is in the process of being generalized to

(7i) before any syllabic suffix only monosyllabic stems have no st0d.

However, matters get worse - or better - according to temperament:

6.3 Non-inflected lexical items

['t:mge:1oa] embede 'office (a post)' ['uhy:1-e] uhyre 'monster'

These syllables with st0d are penultimate in the lexical item and they

should be st0dless. However, the words end in a vowel which is phonetic­ally identical to a semi-productive suffix (infinitive I G I ; in uhyre I G I fuses with the preceding I r I > [ l3 ]) which we have already seen to trigger st0d in a preceding polysyllabic stem. So perhaps principle (7) is sneaking in where it does not really belong, that is within a lexical item. Or perhaps it is the generalization we proposed in (7i) which penetrates the lexical item: before any syllabic suffix only monosyllabic stems have no st0d.

6.4 Non-inflected compound nouns and adjectives

['vi:ng:,m\] vingummi 'wine gum' butgummi alone is without st0d;

['uvil1ja] uvifje 'ill-will' but vi!Je alone is without st0d;

['liwsgk:1oa] livsglade 'joie-de-vivre' but glade alone is without st0d;

['mo:nnsc}ei1 a] morgenstiiie (adj) 'morning quiet' but stifle alone is without st0d.

However, vingummi ends in a vowel, [i], which is phonetically identical to a semi-productive derivative suffix ('-ig' j i j) which, in accordance with (7) and exemplified in 4.1.2.3 above, induces st0d (that is, does not induce non-st0d) in polysyllabic stems.

Uvi!fe, livsgk:ede, morgenstille end in schwa, phonetically identical to the semi-productive infinitive suffix which induces st0d in polysyllabic stems.

Danish st0d - Towards simpler structural principles? 43

What we are witnessing may be a change from a specific morpho­logical parsing, cf. (7), to a mere recognition of morphological complexity which is indiscriminate with regard to the nature of the specific suffix, cf. (7i). From there we move towards a parsing which relies on the iden­tification of certain sounds at the end of the word and their separation from a preceding pseudo-stem, so that:

(7ii) in any word which phonetically resembles a stem + a syllabic suffix, only monosyllabic stems have no st0d.

The net result is that more and more heavy syllables will have st0d, and the need to formulate principles for its absence will diminish. Accordingly, one could argue that the driving force behind such a change is the simplific­ation it entails for the speaker against the concomitant loss in morpho­logical specificity. Simultaneously, the revised st0d principles make the general function of st0d (in the native-like vocabulary) as a signal for heavy syllables e':en more explicit, to the advantage of the speaker and perhaps also the listener.

6.5 Theoretical implications

The changes we have observed do not affect the phonological conditions for st0d: it will still only occur in heavy - or bi-moraic - syllables. Lexical specification for st0d or non-st0d is also largely unaffected. But take the tendencies observed to their logical conclusion - lexical specification apart - then the only heavy syllables without st0d we encounter in the nativevocabulary will be found in disyllables with initial stress. In that restrictedcontext it is still relevant to parse the word morphologically and identify thesuffix. Thus [ 1hu:1sGo] and [1hu:sGo] still demonstrates a surface contrastbetween huset 'the house' (that is ['hu:1s] hus (n) plus the productive definiteI Gd I) and 'housed' (that is ['hu:sG] huse (in£) plus the productive participleI Gd I with I G I + I G I fusing into one I G I). In any other word structure thedistinction between different degrees of productivity in suffixes becomesvoid. However, the crucial aspect of the non-st0d principle is maintained:the segmental string of any word with non-final stress is scanned from theright in order to identify phonological candidates for a putative suffix and itis scanned further left to see whether any syllable(s) precede(s) the pen­ultimate stressed one, in which case the stressed syllable has st0d, providedit is heavy.

The productivity of this purported change in the nature of the st0d­governing principles can be tested in fairly straightforward production ex­neriments. We intend to address this issue in the future.

44 Nina Grnnnum & Hans Basb0ll

7 Conclusion

The principles governing the presence or absence of st0d in heavy syllables may be changing from being expressions of grammatical conditions with a significant distinction between different grammatical categories towards expressing phonological conditions in the word so that only the phono­logical form of suffixes is relevant, not their grammatical content ..

If this is not a passing phenomenon, if it spreads and gains foothold, it will have over-arching consequences for st0d and its functions: only in di­syllables will st0d be truly distinctive and have morphological raison-d'etre. In other contexts it will become predictable from syllable structure and word structure alone and thus independent of the morphological content.

Admittedly, our data are limited and it is too early to do more than conjecture and hypothesize as we have done. We shall have to leave it to phonologists further down the line of generations to complete the picture. Note, however, that whether or not the tendencies towards a change (as depicted in section 6) are carried through to their logical conclusion, st0d _ will still have - as it does to-day - sociolinguistic implications, placing the speaker in a wider language context (cf. section 1); it will still have un­disputed distinctive function (cf. section 3); and its wider communicative function - although less differentiated - as a signal for word structure is maintained.

8 References

Basb0ll, H. (1988): The Modern Danish St0d and Phonological Weight. In: P.M.Bertinetto, M. Lopocarno (eds): Certamen Phonologicum. Proceedings from the Cortona

Phonology Meeting (pp. 119-152). Torino: Rosenberg & Sellier. Basb0ll, H. (1998): Nyt om st0det i modeme rigsdansk - om samspillet mellem

lydstruktur og ordgrammatik. Danske Studier 1998, 33-86. Basb0ll, H. (2003): Prosody, productivity and word structure: the st0d pattern of

Modem Danish. Nordic Journal of Linguistics 26, 5-44. Basb0ll, H. (2005): The Phonology of Danish. Oxford: Oxford University Press. Basb0ll, Hans (2008): St0d, diachrony and the Non-St0d Model. North-Western European

Language Evolution(= NOWELE) 54/55, 147-189. Ejskjrer, I. (1990): St0d and pitch accents in the Danish dialects. Acta Linguistica

Hafniensia 22, 49-76. Fischer-J0rgensen, E. (1987): A phonetic study of the st0d in standard Danish. Annual

&port of the Institute of Phonetics. University of Copenhagen 21, 55-265. Fischer-] 0rgensen, E. (1989a): Phonetic anafysis of the stod in Danish. University of Turku.Fischer-J0rgensen, E. (1989b): Phonetic analysis of the st0d in standard Danish.

Phonetica 46, 1-59. Gr0nnum, N. (1999): Syllables at multiple levels of representation in Danish. Journies

d'itudes Linpuistiaues 1999. 24-29.

Danish st0d- Towards simpler structural principles? 45

Gr0nnum, N. (2009): A Danish phonetically annotated spontaneous speech corpus (DanPASS). Speech Communication 51, 594-603.

Gr0nn=:-, N. and H. Basb0ll (2001a): Consonant length, st0d and morae in standard Darush. Phonetica 58, 230-253.

Gr0nnum, N. and H. Basb0ll (2001b): Consonant length, st0d and morae in Danish. Fonetik 2001, Working Papers, Department of Linguistics, Lund University 49, 46-49.

Grnnn=:-, �- and _H. Basb0ll (2002a): St0d and Length: Acoustic and CognitiveReality. Proceedings of the 1 st International Coeference of Speech Prosotfy, Aix-en-Provence,France, 355-358.

Gr0nn=:-, N. and _H. Basb0ll (20026): St0d and Vowel Length: Acoustic and CognitiveReality? Fonetik 2002, TalMusik Hb'rse4 Quarterfy Progress and Status &port, Stockholm,°Rf!yal Institute of Technology 44, 85-88.

Gr0nnum'.

N. and H. Basb�ll (2003a): Two-phased st0d vowels - a cognitive reality? Fonetik 2003, &ports zn Phonetics. Department of Philosopf?y and Linguistics Ume!zUniversity, Phonum 9, 33-36.

'

Grnnnum, N. and H. Basb0ll (2003b): St0d and Length in Standard Danish: Expe�ents in Laboratory Phonology. Proceedings of the 15 th International Congress ofPhonetic Sczences, Barcelona, Spain, 455-458.

Gr0nnum, N. and H. Basb0ll (2007): Danish St0d: Phonological and Cognitive Issues. In: M.-J. Sole, P. S. Beddor, M. Ohala (eds): Experimental Approaches to Phonology(pp. 192-206). Oxford: Oxford University Press.

Hansen, Aa. (1943): St0det i Dansk. Det KgL Danske Videnskabernes Selskab, Historisk­ftlologiske Meddelelser, XXIX 5. K0benhavn: Munksgaard.

Hombert, J.-M. (1986): Word Games: Some Implications for Analysis of Tone and Other Phonological Constructs. In: J.J. Ohala, J.J. Jaeger (eds): ExperimentalPhonology (pp. 175-186). London: Academic Press.

Hyman, L. (1985): A Theory of Phonological Weight. Dordrecht: Foris. H0ysgaard, J.P. (1743): Concordia res parvre crescunt, eller Anden Prnve af Dansk

Orthographie. K0benhavn. Reprinted in: H. Bertelsen (ed.): (1920/1979): DanskeGrammatikere W (pp. 217-247). K0benhavn: Det Danske Sprog- og Lltteraturselskab.

H0ysgaard, J.P. (1747): Accentuered og Raisonnered Grammatica. K0benhavn. Reprinted in H. Bertelsen (ed.): (1920/1979): Danske Grammatikere W (pp. 249-488). K0benhavn: Det Danske Sprog- og Litteraturselskab.

Kjrerb:ek, L. and H. Basb0ll (2010): Morfologisk udvikling hos danske b0m i alderen 0-10 ar: substantivemes pluralisdannelse. In: H. Vejleskov (ed.): Bornesprog. Fra det12. Nordiske Symposium om Bornesprogsforskning (pp. 15-27). K0benhavn: Forlagetucc.

Kohler, K.J. (1994): Glottal stops and glottalization in German. Data and theory of connected speech processes. Phonetica 51, 38-51.

Kohler, K.J. (2001): P_losive-related glottalization phenomena in read and spontaneouss�eech. A st0d l!l German? In: N. Gr0nnum, J. Rischel (eds): To Honour EliFischer-Jorgensen. Travaux du Cercle Linguistique de Copenhague 31 (pp. 174-211). Copenhagen: Reitzel.

Liberman, A.S. (1982): Gennanic Accentology 1: The Scandinavian Languages. Minneapolis: University of Minnesota Press.

Martinet, A. (1937): La phonologie du mot en danois. Paris: Llbraire C. Klincksieck. Also in Bulletin de la Societe Linguistique de Paris 38: 169-266.

46 Nina Grennum & Hans Basb0ll

Magnus, J. (1554): Historia de omnibus Gothorum Sueonumque regibus. Romae. [Published by 0. Magnus]. 2nd ed. 1617 [Published by Schurer]. r

Rische!, J. (1983): On unit accentuation in Danish - and the distinction between deep and surface phonology. Folia Linguistica 17, 51-97.

Sternberger, J.P. (1984): Length as a suprasegmental: Evidence from speech errors. Language 60, 895-913.

Soderberg, V. (1908): Det Hemming Gadh tilskrifna talet mot danskama. Historiska Studier tillagnade Professor Harald Hjarne pa hans sextioarsdag den 2. maj 1908 af larjungar, 645-674.

Trubetzkoy, N.S. (1935): An/eitung Z!' phonofogischen Beschreibungen. Brno: Edition du Cerc/e Linguistique de Prague. English translation 1968: Introduction to the Principles of Phonological Descriptions. The Hague: Mouton.

Ilse Lehiste (Columbus)

Prosodic allomorphs in the Estonian declension system

1 Introduction

The paper deals with a particular aspect of Estonian grammar - the interaction of prosody and morphology. It is a well known fact of Estonian morphonology that in some nominal paradigms, case is ex­pressed by the prosodic shape of the word. In this paper, I present a classification of Estonian declension on the basis of the use of prosodic case allomorphs.

Estonian nominal inflection involves fourteen cases in the singular and plural, of which the nominative singular, genitive singular, and partitive singular serve as the basis from which the other cases are derived - usually ,with case endings appropriate for each case (for a recentO"\[.erview, cf. Viitso 2003). The nominative, genitive and partitive aresometimes referred to as grammatical cases; the eleven additional cases areconsidered semantic cases. A considerable number of words also have atwelfth semantic case, the fifteenth case, referred to as the short illative oraditive (Viks 1982), and its relevance will become apparent later in thediscussion.

The genitive ends in a vowel, the so-called theme vowel. If the stem already ends in a vowel, there is no additional ending. The partitive can have several forms. If there is a segmental suffix, it is /t/ (for example, pere 'family' -peret, puu 'tree' - puud). There are also numerous paradigms with a zero ending in the partitive, sometimes called the zero allomorph of the case ending (e.g. elu 'life', nominative elu, genitive elu, partitive elu). A problem arises here in defining the function of the final vowel: in the nominative it is part of the word, but in the genitive and partitive it is treated as the stem vowel, in the latter instance being followed by the zero allomorph of the partitive suffix.

There are certain problems with this kind of analysis - the kind that operates with stems and suffixes. The final vowel of elu is part of the lexical identity of the word. If a word ends in a consonant in the nominative case, it acquires a so-called stem vowel in the genitive. If the word already ends in a vowel in the nominative, it keeps the same vowel in the genitive, but here it is either reinterpreted as the stem vowel, or the word acquires a zero allomorph of the genitive ending. Words like elu are identical in the genitive and partitive, but the final vowel in the partitive