A Contrastive Survey of Stress Assignment
in Danish and Norwegian*
Part 1 – Simplex words
Vladimir Naydenov
Статья подчинена цели представить сопоставительное обозрение принципов, определяющих место ударения в так называемом новейшем литературном копенгагенском произношении датского языка и в городском восточном произношении норвежского языка. Автор стремится определить, какие из видимых различий и сходств присущи самим языкам, и какие являются следствием отличающихся друг от друга анализов и теоретических подходов, применявшихся к ним до сих пор. В первой части рассмотрены степени ударения и системные тенденции в фиксации ударении мономорфемных словоформ. Приводятся доводы в поддержку убеждения, что правила, определяющие место ударения, возможно моделировать однотипным способом в обоих языках, и что это осуществимо в рамках метрической фонологии. В то же время указывается на некоторые различия, в особенности касающиеся связи ударения с фонологической длиной.
This is the first part of a two-part article aiming to present a contrastive survey of stress assignment in Advanced Standard Copenhagen Danish and Urban East Norwegian.[1] The goal is to establish which of the apparent differences and similarities are inherent in the languages and which stem from the differences between the analyses and theoretical approaches that have been applied to them in existing descriptions. Part 1 discusses the levels of stress and the patterns of stress in simplex words. It is argued that in both languages, the rules governing stress placement in simplicia can be modelled in similar ways, and that this can be done within the framework of metrical phonology; however, certain differences are also highlighted, especially as regards the connection between stress and quantity.
1. Introduction
Standard Danish and Norwegian Bokmål are very closely related languages. Not only do both belong to the North Germanic branch of the Germanic languages, but in addition Bokmål in fact originated relatively recently as a regional form of Danish influenced by a Norwegian substrate and Norwegian dialects.[2] The phonological and phonetic differences between the two languages are nevertheless striking; indeed, while the written forms are mutually intelligible, the spoken ones present significant difficulties in that respect.[3] However, stress placement is an unexpected exception to this general state of affairs; it seems to display few surface differences, as will be demonstrated in the following. This nearly identical stress placement, viewed against the background of the fairly different phonological systems, within which it occurs, is intriguing and needs an explanation.
The specific varieties to be discussed here are Advanced Standard Copenhagen Danish and Urban East Norwegian. Both of these terms need explanation. The former was coined by Danish phonologist Hans Basbøll (Basbøll 1969, cited in Basbøll 2005:16) to refer to a spoken form of the standard language, which is characteristic of the capital Copenhagen and is neither highly conservative nor associated with the “socially low”. The latter was introduced by Norwegian phonologist Gjert Kristoffersen (Kristoffersen 2000:8-10) to refer to a cluster of quasi-standard urban spoken speech forms in eastern part of Southern Norway, including the capital Oslo; these forms are in turn fairly close to ‘Standard East Norwegian’ (nor. Standard Østnorsk), a variety which reflects formal middle-class urban speech and can be characterized roughly as “spoken Bokmål”.
The present enterprise can be described as a ‘contrastive analysis’ in that two languages are being systematically compared in a certain respect. However, it does not follow strictly any of the methodologies proposed within that tradition. One reason for that is that much of the research in the field known as Contrastive Analysis has been oriented primarily towards applied linguistics and more specifically Second Language Learning. Some authors (James 1989:3-8) seem to reject purely theoretical, non-applied Contrastive Analysis in principle; others (Krzeszowski 1990:9, see also van Els et al. 1984) do recognize it as a possibility. In any case, the methodology proposed and used in most works has been coloured by this general orientation. Few studies of suprasegmental phenomena have been done within that tradition and many of these have been criticized for doing little more than “juxtaposing independent descriptions of data” (Krzeszowski 1990:60) and in particular for not specifying a tertium comparationis (an important concept in the field, denoting the features that the objects to be compared have in common and that serve as a basis of comparison). In this case, the function of tertium comparationis is, hopefully, fulfilled by the very linguistic concepts discussed (segments, syllabic structure, stress as a type of prominence characterizing certain syllables, etcaetera), as the phenomena in question undoubtedly exist in both languages, and are arguably universal (perhaps only as latent constraints, in the case of stress). This is in line with James’ (1989) suggestion that phonological (ibid., p.83) and phonetic (ibid., p.71) universals may be used in this way.
The most recent and prominent descriptions of stress assignment in the languages concerned are Grønnum 1998, 2001 and Basbøll 2005 for Danish and Kristoffersen 2000, Rice 2003 and 2006 for Norwegian[4], and these are the primary descriptive sources used here. These works are written within different theoretical frameworks, and while the adoption of a single theoretic approach is normally one of the preliminary steps towards a contrastive analysis (as pointed out e.g. in James 1989:63, Krzeszowski 1990:107-108), in the present case the very attempt at such a unification will occupy much of the following pages. Kristoffersen’s (2000) description is from the point of view of Lexical Phonology; he has also presented an Optimality Theory (OT) version of the part dealing with simplex words (1999). Rice (1999, 2003, 2006) has developed an alternative OT analysis, also dealing with simplex words, which Kristoffersen (2003:129) has endorsed. Nevertheless, Kristoffersen’s (2000) account still remains the most recent one dealing with all areas of Norwegian stress assignment rather than merely simplex words, so it will be used here as well. As for the framework of Basbøll (2005), it is very original and independent of most current approaches. The work features an innovative model of morphological-and-phonological domains (based on productivity) and strives towards psychological interpretability as well as closeness to phonetics and empirical observation in general.
The rest of the paper is structured as follows. Section 2 of Part 1 deals with the number of stress levels and their expression. Primary, secondary and „tertiary“ stress are described, first for Danish and then for Norwegian. Section 3 discusses stress assignment in simplex words and its relation to quantity. In section 3.1, Basbøll’s (2005) account of Danish is outlined, followed by Kristoffersen’s (2000) and Rice’s (2006) analyses of Norwegian in section 3.2. In section 3.3, Danish is compared to Norwegian. First, some similarities that may not be obvious from the existing descritptions are pointed out (3.3.1). Secondly, it is argued that differences in the quantity system pose difficulties for an identical metrical analysis of the two languages (3.3.2). Finally, an alternative view of quantity and stress in Danish is sketched (3.3.3).
Stress in affixed words and so-called ‘compound stress’ will be discussed in Part 2.
2. The levels of stress and their characteristics
Before moving on to the issue of stress assignment itself, it is necessary specify the number of stress levels that are sufficient to describe the two languages and the means by which they are signalled and distinguished.
2.1. Danish
2.1.1. Primary stress
For Danish, the acoustic characteristic mentioned most frequently and emphatically in connection with primary stress is the presence of a tonal signal associated with the primary stressed syllable (Grønnum 1998:200, 206, Basbøll 2005:330-333).[5] In addition, Danish primary stress is reported to be distinguished from other levels by somewhat greater duration of the syllable (Fischer-Jørgensen 1984, cited in Basbøll 2005:331; Grønnum 1998:200). Grouping these two together, as is commonly done, is justified from a phonetic point of view in that both can be used by listeners as cues for the place of stress. However, it may be somewhat unfortunate in a phonological description, because this prevents one from distinguishing clearly between a a purely phonetic correlate of stress (namely, increased relative duration) and a phonological entity that happens to co-occur with stress (namely, a pitch accent, i.e. a sequence of tones associated with the stressed syllable). In general, the two phonological categories “stress” and “pitch accent” are usually clearly separated in modern linguistic descriptions (see e.g. Ladd 1996:42-43, 46-51; Gussenhoven 2004:13-14, Bruce 1998:97, 100-102). For example, words are known to be deaccented in order to signal information structure in a wide variety of languages, without thereby losing their (primary) stress.[6] Another relevant fact is that most so-called „intonation languages“ allow the speaker to choose between more than one intonational gestures to be associated with the primary stressed syllable, a discrete choice that must be made within the phonology. While it may be that neither of these possibilities exists in Danish[7], this peculiarity can be accounted for synchronically in terms of specific parameter settings or constraint rankings, while the very distinction between the two basic prosodic concepts should be maintained on cross-linguistic grounds. For these reasons, only duration should be regarded as a manifestation of primary stress.
2.1.2. Secondary stress
The issue of secondary stress in Danish causes disagreements even among native speakers and qualified phonologists. For example, Grønnum (1998:205-207) states that all compound members but the primary stressed one are considered to be reduced to secondary stress, and then goes on to express some skepticism about the very phonological reality of this concept in Danish. In her view, alleged secondary stressed syllables usually can’t be distinguished from unstressed ones by any independent criteria apart from the fact that the same morphemes occur with primary stress on these syllables when outside the compound. For example, the only reason why bold in fodbold [‘foðˌb̥ʌlˀd] (“football’) can be identified as secondary-stressed is because it occurs as an independent content word with primary stress (‘[b̥ʌlˀd] “ball’).[8] Even though Grønnum admits that words such as ‘yndlings|ˌteknikker ‘favourite technician’ and ‘yndlings|tekˌnikker ‘favourite techniques’ (which correspond to the independent words ‘teknikker ‘technician’ and tek‘nikker ‘techniques’ respectively) can be distinguished phonetically by the somewhat greater length and slightly more energetic articulation of certain segments in the alleged secondary stressed syllables, she suggests that the second compound members somehow manage to inherit the properties of their primary stressed counterparts without bearing any phonological secondary stress. This seems far-fetched: even if such „inheritance“ effects (rather than regularly assigned secondary stress) are at work here, they still must be generated by the phonology in some way, for example through so-called Output-Output Correspondence constraints (Benua 1997) in the framework of Optimality Theory. But such constraints would need to refer to discrete phonological features, not to a diverse bundle of weak and gradient phonetic effects of the type described by Grønnum. These latter effects must be the phonetic correlate of a single discrete feature that the constraint can refer to, and the most obvious candidate for such a feature is, again, stress. Thus, ironically, it is precisely the elusiveness of the phonetic manifestation of secondary stress that proves its phonological reality.
Unlike Grønnum, Basbøll (2005:333-340) believes that secondary stressed syllables do indeed exist and are distinguished from unstressed syllables by virtue of sharing the primary stressed syllables’ ability to contain vowel length and stød, as in the example ‘b̥ʌlˀd above (stød is a peculiar Danish syllabic prosody with a phonetic realization that includes laryngealization – normally – and glottal stop – sometimes; it is generally transcribed as a glottal stop in IPA). Compound members that have lost vowel length and stød are thus unstressed rather than just secondary stressed. If this is our only criterion, a practical problem arises in that many syllables have neither vowel length nor stød even when primary stressed, so we are unable to tell apart secondary stress and absence of stress in these syllables: e.g. menu [meˈny] ‘menu’ – aftenmenu [ˈɑfd̥n̩me(ˌ)ny] “evening menu’. Yet native speakers appear to perceive secondary stress even on such syllables, as we shall see below. As mentioned above, besides the phonological criterion proposed by Basbøll, secondary stressed syllables have been claimed to occupy an intermediate position between primary and absence of stress in terms of syllable duration, and to have higher intensity than unstressed ones.
Finally, at the other end of the spectrum of opinions, Fischer-Jørgensen 1984 (cited in Basbøll 2005:333-340) believes to be able to distinguish not one but two degrees of secondary stress.
2.1.3. Absence of stress
Absence of stress (termed “tertiary stress’ by Basbøll) hardly needs much commenting. It is claimed to have smaller duration and intensity than higher stress levels. In Basbøll’s system, it is characterized by the syllables’ inability to have vowel length or stød, but Grønnum (2001:237) doesn’t seem to accept that word forms like sofaen [ˈsoː(ˌ)fæːˀn̩] (‘sofa’, def. sing.) and hyppige [ˈhy(ˌ)b̥iː(ˀ)i] (‘frequent’, adj. def./plur.), have secondary stress on their long and/or stød-bearing syllables. The schwa vowel and syllabic sonorants can only occur in unstressed syllables.[9]
2.2. Norwegian
2.2.1. Primary stress
In Norwegian, the situation is similar in many ways. According to Kristoffersen (2000:141), the unique characteristic of Norwegian primary stressed syllables is their ability to sustain tonal accent. In this language, this means that they will serve as docking points for one of two melodies – commonly referred to as accent 1 and accent 2 – the choice between the two being partly lexically and morphologically conditioned.[10] Again, as in Danish, a remark is called for: even if we assume that primary stress really lacks its own phonetic correlates and is only recognizable by virtue of its obligatory co-occurrence with pitch accent, this does not mean that the latter is a merely a „manifestation“ of the former (nor does Kristoffersen suggest that). This caution is even more pertinent in Norwegian, given the fact that deaccenting is well-attested as a regular device of Norwegian intonation. Of course, it is conceivable that deaccenting in Norwegian does constitute a reduction of primary stress to secondary, or that primary stress has no obligatory signalling of any kind. However, the best option seems to be to look for other phonetic correlates of this stress level.
2.2.2. Secondary stress
Secondary stress is usually attributed to the non-primary stressed members of compounds in Norwegian. Like primary stressed syllables, secondary stressed ones are characterized by the possibility, and indeed the requirement, that they should be heavy. This entails the presence of either a long vowel or a coda consonant, often an ambisyllabic geminate (to be discussed in more detail below). An example is [ˈfuːtˌbɑlːˌlaːg] “football team’, a three-root compound.
Kristoffersen (2000) regards the above as ‘strong secondary stress’. He also argues (p.141, 162-165) for an additional level in Norwegian, namely weak secondary stress in the initial syllables of words with non-initial stress consisting of more than three syllables such as delikat [(ˌ)dɛlɪ1khɑːt] ‘delicate’. Although admittedly subtle and never proved experimentally, this category is expressed by a slight lengthening of the post-vocalic consonant and by the realization of /e/ as [ɛ] as opposed to [ə]. To address this very briefly, it seems to stem from a requirement for a left-aligned trochee (moraic or syllabic) and is similar to the one attested in German (Féry 1996:63). It has not been found in Danish as yet and will not be dealt with here any further, although the question whether a similar generalization obtains in that language deserves further investigation (cf. telefon ‘telephone’, which at least the ODS transcribes as [tselə’foːˀn], with a difference in vowel quality between the two unstressed syllables).
2.2.3. Absence of stress
Unstressed sylables are unable to sustain long vowels and geminates. As in Danish, schwa may only occur in such syllables.
As a whole, the similarities between the two languages are obvious. The most conspicuous differences stem from the fact that stressed syllables in Danish are not subjected to equally strict weight requirements as those in Norwegian, hence the boundary between secondary and absence of stress is much more fuzzy and contentious in the former language than in the latter. It is a curious fact that although Norwegian tonal accent is historically related to Danish stød, the synchronic point of view actually approximates it to another tonal phenomenon, Danish pitch accent (as both are associated with primary stress only), whereas stød behaves like vowel length (as both are realized entirely within the syllable and are possible under both primary and secondary stress, but not in unstressed syllables).[11]
3. Stress in simplex words.[12] The relation between stress and quantity
3.1. Danish
Basbøll (most recently in 2005:395-400) accounts for stress assignment in Danish simplex words in the following way, which is accepted also by Grønnum (2001:200-202). First of all, he splits them into two lexical classes, one composed of French loanwords and one comprising all the rest. French loanwords (e.g. me‘ny ‘menu’) invariably have final stress (schwa being exempted from this, of course: assi‘stance [asiˈsd̥ansə] ‘assistance’). The other class is characterized by a more complex rule. If the word contains one or more long vowels, the first one receives the (primary) stress (if there are other long vowels than the primary stressed one, they automatically receive secondary stress in order to preserve their length, according to Basbøll). This can be exemplified with the words sofa [ˈsoːfa] (‘sofa’) and paradis [‘phɑːɑˌd̥iːˀs] (‘paradise’).[13] If the word does not contain any long vowels, stress falls on the last vowel followed by a consonant, e.g. kalif [khaˈlif] (‘caliph’), salmonella [salmoˈnɛla] (‘salmonella’). Finally, since many words don’t seem to obey these default rules, two groups of words with lexically pre-specified stress are posited: those with exceptional stress on the antepenult (Mexico [ˈmɛɡ̊sikho], ‘Mexico’ vs expected penultimate stress) and those with exceptional stress on the penult (turban [ˈtsuɐ̯b̥an], ‘turban’ vs expected final stress).
Unlike the above description of Danish, recent analyses of Norwegian stress are all cast within the framework of metrical theory (e.g. Hayes 1995). As Kager (2007) puts it, this is ‘a theory whose central assumption is that stress is a relational property, represented by prominence relations between constituents in hierarchical structures’. Without digging too deep into the formalisms of this approach, in a simplex word it essentially entails that one or more metrical feet are constructed and stress is assigned to the head of one of these feet according to certain rules/constraints.[14] Furthermore, the analyses discussed here use the mora (a unit of weight and/or length).
3.2. Norwegian
3.2.1. A rule-based account
Kristoffersen (1999, 2000:147-162) starts by formulating a chicken-and-egg problem concerning the relationship between stress and quantity in Norwegian: on the one hand, a syllable containing a long segment (vowel or consonant) must always be stressed; on the other hand, a stressed syllable must always be heavy (bimoraic), i.e. it must contain either a long vowel or a coda consonant (often a geminate). For example, [2bɑːnə] ‘track, path’, [2bɑnːə] ‘curse (inf.)’ and [2bɑndə] ‘gang’ are all legitimate words, but *[2bɑnə] and *[2bɑːnːə] are impossible.[15] He concludes that quantity (of vowels and consonants) has to be secondary, because stress can be predicted with a high degree of accuracy on different grounds, as explained below. Obviously, an additional chicken-and-egg problem is the one of vowel length vs consonant length: in a stressed syllable, a long vowel entails a short consonant and vice versa. Since Kristoffersen has established that quantity itself must be secondary, it can’t be specified in the underlying form, so neither vowel length nor consonant length can be primary; instead, he proposes that morphemes are divided into two lexical classes (the [2bɑːnə] type and the [2bɑnːə] type), each with its own way of making the syllable bimoraic when stressed.
Once quantity is excluded from the picture, Kristoffersen establishes that both stress in words belonging to the native core such as gate [2gɑːtə] (‘street’) and stress in polysyllabic loanwords such as assistance [ɑsɪ1stɑŋsə] (‘assistance’) and trafikk [trɑ1fɪk] (‘traffic’) can be predicted by the same rule, one that constructs a moraic trochee at the right edge of the word, counting each vowel and each coda consonant as moraic.[16],[17] In other words, Norwegian typically stresses the last closed syllable, else the next-to-the-last open syllable; a very common rule, which also holds true e.g. of English (although the final syllable of English nouns is invisible to the rule). This means that the underlying forms are, say, /bɑne/1 , /bɑne/2, /trafik/, without quantity; the initial footing is, accordingly (ba.ne), (ba.ne), tra.(fik). Stress is assigned to the the head of the trochaic foot, i.e. to the syllable containing the second mora counting from the left, producing /ˈbɑne/1 , /ˈbɑne/2, /traˈfik/. Only afterwards is the syllable made heavy by lengthening some segment if necessary, giving rise to /ˈbɑːne/, /ˈbɑnːe/, and /traˈfik/ respectively. Like Basbøll for Danish, Kristoffersen is forced to assume lexical exceptions to account for all words. Antepenultimate stress is explained by the final syllable being lexically specified as “extraprosodic” (or “extrametrical”), a well-known device within metrical theory. Thus, in words like Portugal [1phɔʈːʉgɑl] (‘Portugal’) and turban [1thʉrban] (‘turban’) (underlying /portugɑl/, /turban/), the last syllable is invisible for metrical structure and the foot is constructed on the segmental material to its left. Furthermore, there are words with unexpected final stress such as orkidé [ɔrki1deː] (‘orchid’); these require lexical prespecification of stress, pure and simple.
3.2.2. A constraint-based account
Rice (2003, 2005, see also Rice 1999 for an earlier version) takes a radical Optimality Theoretic (OT) approach to the issue. An OT grammar (Prince & Smolensky 1993) represents a language-specific ranking of a set of universal constraints on the phonological output, and the optimal candidate is determined on the basis of violating as few and as low-ranking constraints as possible. Since it is assumed that there are no constraints on the input (i.e. on the underlying forms stored in the lexicon), a logical conclusion is that one should attempt to determine not the underlying forms per se, but rather the particular OT grammar that would always produce the type of forms that we know to be legitimate in this particular language, regardless of the properties of the input. Hence, Rice doesn’t attempt to solve Kristoffersen’s chicken-and-egg problem in a unitary way, but rather assumes that the final output results from a complex interplay between constraints, some of which (AlignRight, FootBinarity) favour a moraic trochee at the right edge, while others (MaxLink-(μ)[seg]) act to preserve underlying quantity. Besides being in line with current trends in phonology, this view is attractive in that it restores the role of underlying quantity, which is clearly a more natural solution to the quantity contrast than grouping each and every word into one of two lexical classes.
Unfortunately, this change doesn’t seem sufficient to incorporate the typologically appealing intuitions of Kristoffersen’s approach in a standard Optimality Theoretic analysis. A crucial feature of Kristoffersen’s description is that the bimoraic trochees it posits are never ‘real’ on the surface in penultimate-stressed words: /ˈbɑne/, or more precisely /ˈ(ba.ne)/, simply does not exist on the surface (because either /n/ or /a/ must be lengthened). Now, unlike Lexical Phonology, the standard form of Optimality Theory does not allow intermediate levels between the underlying form and the surface form. All constraints apply simultaneously, both footing and lengthening take place at the same time, and there is simply no point at which the foot can be constructed in the way Kristoffersen argues for. While (ˈban).ne or (ˈbaː).ne would also be bimoraic trochees, it would remain unexplicable why *ba.(ˈneː) isn’t preferred over them, as it obviously fares better with respect to AlignRight.[18] Hence, another key element that distinguishes Rice’s account from Kristoffersen’s is the use of a NonFinality constraint, which bans stress on the final syllable and thus favours (ˈbaː).ne over *ba.(ˈneː). Most words with apparent final stress (trafikk [trɑ1fɪkː] /trafikk/, ‘traffic’, and tomat [tɔ1mɑːt], /tomat/, ‘tomato’) are analysed as not violating the constraint, because the final consonant is actually counted as the onset of an abstract “empty syllable”[19]. Thus, the tendency for final open syllables to be unstressed is explained in a different way from Kristoffersen’s, whereas a requirement for a bimoraic trochee does exist, but is satisfied merely by the obligatory length of the stressed syllable.
The OT grammar employed by Rice is Stress-to-Weight, FootBinarity >> MaxLink-(μ)[seg] >> Weight-to-Stress >> NonFinality, NoCoda >> AlignRight. These constraints require, respectively: that a stressed syllable should be heavy, that a foot should consist of two constitutents, that underlying morae should be preserved, that a heavy syllable should be stressed, that a stressed syllable shouldn’t be final in the word, that a syllable shouldn’t have a coda, and that the stressed syllable should be as far to the right in the word as possible. Without delving further into the depths of OT formalisms, the effect of this constraint ranking can be summarized as follows. We can regard one mora per vowel and no moraic consonants as the default situation for the input; in the output, this produces penultimate stress and a long vowel in words ending in an open syllable (bane, /bɑμneμ/ or /bɑμμneμ/ > /ˈ(baː).ne/’track, path’) and final stress and a long vowel in words ending in a closed syllable (to1mat, /toμmaμt/ or /toμmaμμt/ > /to.ˈ(maː).t/’tomato’). The quantity pattern that deviates from this default results from the underlying form having a mora linked to a coda consonant (banne /bɑμnμeμ/ > /ˈ(bɑn).ne/ ‘to curse’, trafikk /traμfiμkμ/ > /tra.ˈ(fik).k/ ‘trafic’). The stress patterns that deviate from the default are the result of underlying forms having an extra mora linked either to the vowel or to the coda consonant of the syllable that ends up with stress (orkidé /oμrkiμdeμμ/ > /or.ki.ˈ(deː)/, eddik ‘vinegar’ /edμiμk/ > /ˈ(ed.)dik/, kokos ‘cocoa [nut]’ /koμμkoμs/ > /ˈ(koː.)kos/; in all these cases, stress is placed on a given syllable in order to preserve underlying quantity). The type with a consonant cluster in the stressed syllable ([2lɪstə], ‘list’) requires (in cases when the cluster is a possible onset) underlying moraicity of the consonant /liμsμteμ/ > /ˈlis.te/), otherwise it would produce a form with a long vowel (this latter, „default“ type actually does occur, but only in a very limited set of words, which makes this one of the more counter-intuitive parts of the theory).
In fact, this analysis is partly similar to the one commonly advocated for English (Pater 1995:16-19, Alderete 1999:122, also e.g. Roca and Johnson 1999). English, while using a bimoraic trochee, has automatic final syllable extrametricality in nouns (agen<da>, brevi<ty>), expressed in OT by a NonFinality constraint. So does Norwegian in Rice’s analysis (ba<ne>), except for the fact that the excluded final syllable has to be empty to derive final stress in closed syllables (toma<t_>). The empty final syllable is somewhat unappealling because of its abstractness, although it has been proposed for Dutch as well (by van Oostendorp 2000, necessitated by the use of syllabic trochees only in that analysis) and for many other languages previously. Making the final consonant obligatorily extrasyllabic as suggested for German in Caratini (2007) would produce similar quantity patterns, but that would make penultimate stress the default even for words with closed final syllables (as in Frid’s 2001 account of Swedish), whereas in fact they are statistically far less common.
3.3. How different is Danish?
The above accounts of Norwegian have the advantage of being based on widely attested principles that have been posited for a variety of languages. This holds true of metrical theory, which has been developed with the ambition to constitute part of Universal Grammar, and even more so to Optimality Theory, in which all constraints on the phonological output are universal. The cited analysis of Danish, in contrast, is more idiosyncratic. Yet, despite the vastly different analyses, the two languages have almost always the same stress in simplex words, which is easily explained diachronically, but remains to be accounted for synchronically. The question naturally arises whether the models used for Norwegian can’t be applied to Danish as well.
3.3.1. Similarities
Danish is more similar to Norwegian than the descriptions suggest. Let us recall that the default rule proposed for Danish is, essentially, ‘stress the first long vowel, else the last vowel followed by a consonant’. This would seem to be case of what has been called “conflicting directionality”, a pattern that is as crosslinguistically rare as it is puzzling and troublesome from an OT perspective (see e.g. Zoll 1996:124-165 for an attempt to explain it and Gordon 2000 for an attempt to argue that such patterns do not exist at all). However, such an analysis is not unavoidable in the case of Danish. The first part of the rule is actually meant to account for two groups of words: words with more than one long vowel such as paradis [‘phɑːɑˌd̥iːˀs] (they are relatively few), and words with a single long vowel such as sofa [ˈsoːfa] (they represent the overwhelming majority of the relevant lexemes). The [‘phɑːɑˌd̥iːˀs] type corresponds to what is known in Norwegian and Swedish phonology as “formal compounds”: despite being morphologically simplex, these words have precisely the same prosodic pattern as real compounds (to be discussed in section 4.2), hence the Norwegian analyses don’t treat them together with “normal” simplex words.[20] In fact, there is no reason to single out long vowels; in both Norwegian and Danish, the same rule applies even when no long vowels are involved: cf. Danish løjtnant [‘lʌɪ̯d̥ˌnanˀd̥] ‘lieutenant’, boykot [‘b̥ʌɪ̯ˌkhʌd̥] ‘boycot’ (the examples, along with the secondary stresses in their transcription, are taken from a monolingual Danish dictionary, the Ordbog over det danske Sprog, and correspond to similar “prosodic compounds” in Swedish). Once we have excluded (temporarily) the formal compounds, little reason remains for the separate treatment of words containing long vowels: sofa [ˈsoːfa] would be assigned stress correctly if subjected to the second part of the rule just as salmonella [salmoˈnɛla] is. As for the second part of the rule, it essentially ensures that stress is assigned to the last syllable if closed (kalif [khaˈlif]), else to the next-to-the-last syllable (salmonella, [salmoˈnɛla]). Precisely the same result is achieved both by the default rule in Kristoffersen’s (2000) rule-ordering moraic account (where stress is assigned to a bimoraic foot, constructed at the right edge of the word) and by Rice’s (2003, 2005) OT account involving a constraint against final stress and an abstract final syllable.
It must be noted that morae are employed by Basbøll (2005:265-293 and passim) for a purpose very different from predicting stress: since syllables containing long vowels and syllables ending in a sonorant consonant are the only two syllable types that permit the prosody known as “stød”, he argues that these and only these two syllable types are bimoraic. Non-sonorant consonants, on the other hand, are not moraic, because they do not admit stød. While similar moraicity principles have been posited for certain other languages (van der Hulst 1999b:12, Hayes 1995:297), little evidence of this type of moraic structure exists in Danish outside of the stød, so this seems to be a case of circular proof. Furthermore, the best-known and arguably most salient feature of the stød constitutes a laryngealization of (roughly) the second half of the syllable (Fischer-Jørgensen 1989), so there is a non-abstract, phonetic reason why it shouldn’t occur in a syllable with a voiceless coda.[21] Thus, the benefits of this use of the mora concept do not seem to outweigh the advantages of its more widespread use to predict weight-sensitive stress.[22]
3.3.2. Differences
Despite all of the above, there are strong arguments against adopting either of the ‘Norwegian’ analyses in its entirety. The Stress-to-Weight principle, so obvious in Norwegian and particularly essential to Rice’s analyses, is difficult to observe in Danish. While Norwegian only allows the types [2bɑːnə] and [2bɑnːə] in stressed syllables, the corresponding Danish lexemes surface (in their most distinct pronunciation, without deletion of schwa) as [ˈb̥æːnə] and [ˈb̥anə]. As manifested in the second word (spelled bande), Danish has no (surface) consonant length.[23] This may be remedied by assuming that the /n/ in this word (and the /f/ in kalif [khaˈlif]) is in fact an underlying or ‘virtual’ geminate, although not pronounced as such, along the lines of Caratini’s (2007) proposal for German and van der Hulst’s (1985) for Dutch. A similar solution is that it is “ambisyllabic”, i.e. linked to both syllables and hence providing the first one with a coda, making it closed (advocated for German e.g. by Hall 2000). It could be said then that ambisyllabic consonants are pronounced as geminates in Norwegian, but not in Danish. Certainly the fact that stød, which is indisputably a property of the stressed syllable only, is pronounced simultaneously with the post-vocalic consonant in surface ˈCVCʔVC0 sequences, e.g. haller [ˈhalʔɐ] ‘hall (plur.)’, suggests that the post-vocalic consonant at least partly belongs to that syllable (although the complex restrictions on stød distribution make it difficult to check whether this is valid for all similar sequences). There are also various other facts that can be perceived as evidence for ambisyllabic consonants in Danish (see Borowski, Itô and Mester 1984), but the patterns described by these authors are found in consonants not only after short, but also after long vowels (in words like ‘bande, but also words like ‘bane), so they don’t seem to be compatible with a Stress-to-Weight interpretation (as observed also by Riad 1995:180). Furthermore, it is a telling coincidence that all of above-mentioned facts concern only words ending in a syllable with a schwa nucleus – words like ‘haller and ‘bande, but not ‘salmonella (more on this below).
The Stress-to-Weight principle is dealt another blow in Danish by the existence of words such as nu [‘nu] ‘now’ (historically, the shortness is generalized on the basis of the pronunciation in unstressed position) and menu [me’ny] ‘menu’ (historically, the shortness is to be explained with the French origin of the word). In contrast not only to the Norwegian equivalents ([‘nʉː], [mɛ’nyː]), but even to the German ones, the vowel is not lengthened in Danish. Some type of lexically specified shortness (as opposed both to lexically specified length and to the default lack of lexically specified quantity) may be conceivable, but does not seem to fit well in the frameworks under discussion.
The final problem to be mentioned here is that according to the „Norwegian“ account, the long vowel in Danish words such as hus [‘huːˀs] (‘house’) is an automatic result of the extrasyllabic status of the final consonant (/hu<s>/), which makes the stressed syllable open and causes its vowel to lengthen. But there is already a group of words assumed to have this syllabic structure, the ven type: in words such as ven [‘vɛn] (‘friend’), the /n/ is considered to be extrasyllabic, because bimoraic syllables normally receive stød, as exemplified in pen (‘pen’) [‘phɛnˀ] (Basbøll 2006:280, passim). Yet no automatic lengthening occurs in ven [‘vɛn].
3.3.3. A possible alternative
The facts enumerated in the previous section suggest that Rice’s description of Norwegian in its present form is unsuitable for Danish because of the apparent weakness of Stress-to-Weight in that language. And, ironically, this very peculiarity may give an advantage of sorts to Kristoffersen’s description. At least in Danish words without long vowels such as bande [ˈb̥anə], a right-aligned bimoraic trochee as in (ˈba.ne) seems reasonably realistic already on the surface. As for the alternative bane [ˈb̥æːnə], it can be explained with lexically specified vowel length as posited by Basbøll 2005 (as opposed to the underlying consonant length postulated by Rice 2005 for Norwegian) on the same grounds, on which Rice accepts underlying length as a factor. That would spare us the disadvantages connected with the approaches that seek to derive vowel quantity patterns from phonetically unrealistic “virtual geminates” and/or ambisyllabicity, in addition to avoiding the postulation of empty syllables. This is also the position taken on Danish by Riad (1992:295,329, 1995).
But unlike the virtual geminate / ambisyllabicity approaches, which postulate some kind of synchronic Open Syllable Lengthening, this one needs a separate explanation for the virtual absence of long vowels in front of consonant clusters within morphemes: for instance lampe [‘lɑmb̥ə]) ‘lamp’ is possible, but not *[‘læːmb̥ə]. Morpheme-finally, a similar restriction obtains, although one “additional” consonant is allowed: hus [‘huːˀs] (‘house’) and telefon [tselə’foːˀn] (‘telephone’) are legitimate monomorphemic words, but not *[‘huːˀsb̥] and *[tselə’foːˀnɡ̊]. An Open Syllable Lengthening account (Kristoffersen’s 2000:129) predicts these facts automatically; but since the alternative under discussion relies on predominantly lexical vowel length instead, the most obvious remaining response is to appeal to a high-ranking constraint banning overlong syllables, which necessitates the shortening of any underlying long vowels under these conditions. As for the difference between morpheme-final and morpheme-internal restrictions, one has to conclude that the above constraint, in combination with faithfulness to underlying vowel length, forces final consonant extraprosodicity on the output: hu<s> [‘huːˀs], telefo<n> [tselə’foːˀn]. The assumption that vowel length is underlying also allows one to explain the difference between hu<s> [‘huːˀs] and ve<n> [‘vɛn], which was mentioned in the preceding section. While Norwegian clearly requires that a stressed syllable should be precisely bimoraic, and thus neither short (monomoraic) nor overlong (trimoraic), Danish apparently has only the latter, „upper“ limit.
At first sight, an additional problem for all the metrical accounts would appear to be the fact that Basbøll (2005:259-260) argues for a much more peculiar syllabification principle than the one assumed for Norwegian; and since those accounts rely crucially on the distinction between open and closed syllables, this could make their predictions for Danish different from those for Norwegian. In Basbøll’s syllabification procedure, which is primarily grounded in the patterns of consonant lenition in Danish, [ˈsoːfa] and [salmoˈnɛla] are syllabified as /soː.fa/ and /sal.mo.ne.la/, as expected; however, a consonant preceding schwa is counted not as belonging not to the onset of the schwa syllable, but to the coda of the preceding syllable, [ˈb̥æːnə] and [ˈb̥anə] being /baːn. ə/ and /ban.ə/, respectively. This does produce many more underlyingly closed syllables than are assumed for Norwegian. While one could, in principle, look for other ways to explain the consonant lenition patterns (say, assuming in line with the historical causes that schwa syllables have an exceptionally weak, fourth degree of stress – compare e.g. the metrical grid proposed by Uhmann 1991:176 for German – so that their onsets are subjected to the same restrictions as the codas of normal syllables), this is not strictly necessary. The syllabification proposed by Basbøll is not fatal for the metrical accounts: after all, they would simply predict all these closed syllables to attract stress, and syllables preceding schwa are indeed, as a rule, stressed in Danish (unless they have schwa as a nucleus themselves).
For the description of primary stress to be exhaustive, it is necessary to mention a characteristic East Norwegian stress pattern or “retraction”; monomorphemic words that have non-initial primary stress in the formal standard language (e.g. orkidé [ɔrki1deː]) may receive primary stress on the first syllable, and secondary stress on the syllable that “should” have had the primary stress. This produces a formal compound (*[2ɔrkiˌdeː]). Essentially, this is no different from the process that produced Danish løjtnant [‘lʌɪ̯d̥ˌnanˀd̥], except that the Norwegian counterpart is still synchronically active, albeit very variable and sociolinguistically marked.
4. Conclusions
While the final conclusions of this survey will only be appropriate in Part 2, after complex words have been discussed, it is possible to make some preliminary observations already here. The stress placement in simplex words is largely the same. However, there are differences in the overall effect of stress in that only Norwegian has surface geminate consonants in stressed syllables, while only Danish has stressed monomoraic syllables. This suggests a difference in underlying quantity and hence a somewhat different mechanism in stress assignment in Danish, even though it leads to the more or less same surface results.
R e f e r e n c e s
Basbøll 1969: H. B a s b ø l l. The phoneme system of advanced Standard Copenhagen. Annual Report of the Institute of Phonetics, University of Copenhagen 3, pp. 33-54.
Basbøll 2005: H. B a s b ø l l. The Phonology of Danish. Oxford University Press.
Behne et al 1998: D. B e h n e, P. E. C z i g l e r, K. P. H. S u l l i v a n. Perceived Swedish vowel quantity: effects of postvocalic consonant duration. Phonum 6, 91–97. Available online http://www.ling.umu.se/publications/articles/phonum06/phonum06.7.pdf
Benua 1997 [1998]: L. B e n u a. Transderivational Identity: Phonological Relations between Words. Doctoral dissertation, University of Massachusetts, Amherst. [Amherst, MA: Graduate Linguistic Student Association, ROA-259-0498]
Borowski, Itô, Mester 1984: T. B o r o w s k i, J. I t ô, A. M e s t e r. The Formal Representation of Ambisyllabicity: Evidence from Danish. Proceedings of the North Eastern Linguistics Society (NELS) 14, 34-48.
Caratini 2007. E. C a r a t i n i. There is no ambisyllabicity (in German), to appear ‘Proceedings of ConSOLE XIV ‘, S. Blaho, E. Schoorlemmer & L. Vicente (Eds.).
van Els et al. 1984: T. v a n E l s. Applied Linguistics and the Learning and Teaching of Foreign Languages. London: Edward Arnold.
Féry 1996: C. F é r y. “German Foot and Word Stress in OT.’ Tromsø University Working Papers on Language and Linguistics 24,1996, 63-96.
Fischer-Jørgensen 1984: E. F i s c h e r – J ø r g e n s e n. The acoustic manifestation of stress in Danish with particular reference to the reduction of stress in compounds. Annual Reports of the Institute of Phonetics, 18,University of Copenhagen ,1984, 45-161.
Fischer-Jørgensen 1989: E. F i s c h e r – J ø r g e n s e n. Phonetic analysis of the stød in Standard Danish. Phonetica 46, 1989, 1-59.
Frid 2001: J. F r i d. Swedish word stress in optimality theory. Working Papers 48. Lund: Lund University, Dept. of Linguistics, 2001, 25–41.
Gordon 1999: M. G o r d o n. Syllable Weight: Phonetics, Phonology, and Typology. Ph.D. dissertation, University of California, Los Angeles, 1999.
Gordon 2000: M. G o r d o n. Re-examining default-to-opposite stress. Proceedings of the annual meeting of the Berkeley Linguistics Society 26. Berkeley, CA: Berkeley Linguistics Society, 2000.
Gordon 2002: M. G o r d o n. A phonetically-driven account of syllable weight, 2002, Language 78, 2002, 51-80.
Grønnum 1998, 2001. N. G r ø n n u m. Fonetik og fonologi, Almen og Dansk. København: Akademisk Forlag, 2001.
Gussenhoven 2004: C. G u s s e n h o v e n. The phonology of tone and intonation. Cambridge: Cambridge University Press, 2004.
Hall 2000: T. A. H a l l. Phonologie: eine Einführung. Berlin: W. de Gruyter, 2000.
Hayes 1995: B. H a y e s. Metrical stress theory. Chicago: Univ. of Chicago Press, 1995.
Hulst 1985: H. v a n d e r H u l s t. Ambisyllabicity in Dutch. H. Bennis & F. Beukma (Eds.), Linguistics in the Netherlands 1985, pp. 57-67. Available online: http://www.dbnl.org/tekst/huls009ambi01/
Hulst 1999: H. v a n d e r H u l s t ( e d. ). Word prosodic systems in the languages of Europe. Mouton de Gruyter, Berlin & New York, 1999.
Hulst 1999b: H. v a n d e r H u l s t. Word accent. P. 3-116. In Hulst, Harry van der (1999) (ed.). Word prosodic systems in the languages of Europe. Mouton de Gruyter, Berlin & New York, 1999, 3-116.
James 1989: C. J a m e s. Contrastive Analysis. Harlow, Essex: Longman. First ed. 1980.
Johnsen 2008: O. E. J o h n s e n. Konstrainter og analogier i det norske trykksystemet. Mastergradsoppgave. Universitetet i Tromsø. Available online: http://www.ub.uit.no/munin/bitstream/10037/1531/1/thesis.pdf
Kager 2007: R. K a g e r. Feet and metrical stress. In Paul de Lacy (ed.) The Cambridge Handbook of Phonology. Cambridge University Press, ch.9, 2007.
Kristoffersen 1999: G. K r i s t o f f e r s e n. Quantity in Norwegian syllable structure. In: van der Hulst, H., Ritter, N. (Eds.), The Syllable. Facts and Views. Mounton de Gruyter, Berlin, 1999, 631–650.
Kristoffersen 2000: G. K r i s t o f f e r s e n. The Phonology of Norwegian. Oxford University Press, 2000.
Kristoffersen 2003: G. K r i s t o f f e r s e n. Norsk prosodi. Manuscript. Nordisk institutt, Universitetet i Bergen, 2003.
Krzeszowski 1990: T. P. K r z e s z o w s k i. Contrasting Languages: The Scope of Contrastive Linguistics. Berlin: Mouton de Gruyter, 1990.
Lorentz 1996: O. L o r e n t z. Length and correspondence in Scandinavian. I: Nordlyd 24, 111-128. Available online: http://www.hum.uit.no/tidsskrifter/nordlyd/Nordlyd24/Lorentz.pdf
Lunden 2006: A. L u n d e n. Weight, Final Lengthening and Stress: A Phonetical and Phonological Case Study of Norwegian. PhD. Santa Cruz: University of California. Available online: http://people.ucsc.edu/~lunden/LundenDiss.pdf
ODS: Ordbog over det danske Sprog – dansk i perioden 1700-1950. Available online: http://ordnet.dk/ods/.
van Oostendorp 2000: M. v a n O o s t e n d o r p. Phonological Projection. Berlin/New York: Mouton de Gruyter, 2000.
van Oostendorp 2002: M. v a n O o s t e n d o r p. The Phonological and Morphological Status of the Prosodic Word Adjunct. In Resolving Conflicts in Grammars: Optimality Theory in Syntax, Morphology, and Phonology. G. Fanselow and C. Féry (eds.), Hamburg: Buske, 2002, 209-35.
Pater 1995: J. P a t e r. On the Non-Uniformity of Weight-to-Stress and Stress Preservation Effects in English. Ms., McGill University, 1995.
Prince, Smolensky 1993: A. P r i n c e, P. S m o l e n s k y. Optimality Theory: Constraint Interaction in Generative Grammar. Rutgers University Center for Cognitive Science Technical Report 2, 1993.
Riad 1992: T. R i a d. Structures in Germanic Prosody. A diachronic study with special reference to the Nordic languages. Diss. Stockholm University, 1992.
Riad 1995: T. R i a d. The Germanic quantity shift. A typology. Amsterdamer Beiträge zur älteren Germanistik 42 (Quantitätsproblematik und Metrik. Greifswalder symposion zur germanischen Grammatik), 1995.
Riad 2000: T. R i a d. The origin of Danish stød. In: Aditi Lahiri (ed.) Analogy, Levelling and Markedness. Principles of change in phonology and morphology. Berlin/New York, Mouton de Gruyter, 2000, 261–300.
Riad 2003a: T. R i a d. Distribution of tonal accent in Scandinavian morphology. Handout at the 1st International Workshop on Franconian Tone Accents, Leiden, June 13-14, 2003.
Riad 2003b: T. R i a d. Diachrony of the Scandinavian accent typology. In: Fikkert, Paula and Haike Jacobs (utg.) (2003). Development in Prosodic Systems. Berlin, New York: Mouton de Gruyter, 2003, 91-144.
Riad 2008: T. R i a d. Prosodi i svenskans ordbildning. Manuscript. Stockholm university, 2008.
Riad, Segerup 2008: T. R i a d, M. S e g e r u p. Phonological association of tone. Phonetic implications in West Swedish and East Norwegian. Proceedings Fonetik, Göteborg, 2008, 93-96.
Rice 1999: C. R i c e. Norwegian. In H. van der Hulst (ed.) (1999). Word prosodic systems in the languages of Europe. Berlin: Mouton de Gruyter, 1999, 545-53.
Rice 2003: C. R i c e. Norwegian quantity and the richness of the base. Manuscript. University of Tromsø, available at http://www.hum.uit.no/a/rice.
Rice 2006: C. R i c e. Norwegian stress and quantity: The implications of loanwords. Lingua 116, 2006, 1171–1194.
Roca, Johnson 1999: I. R o c a, W. J o h n s o n. A Course in Phonology. Blackwell Publishing, 1999.
Zoll 1996: C. C. Z o l l. Parsing Below the Segment in a Constraint Based Framework,1996.
* I would like to thank Tomas Riad and Hristo Stamenov, who read preliminary versions of this paper and offered numerous constructive suggestions for improvement. Thanks are also due to Snezhana Dimitrova and Vladimir Zhobov, who read an early draft of the paper and provided feedback and valuable advice. All errors are mine. Some of the research reflected in the paper was made possible by a grant from the Swedish Institute.
[1]The terms will be explained below. In much of the following text, the designations ‘Danish’ (abbreviated as dan.) and ‘Norwegian’ (abbreviated as nor.) are used for short.
[2] Norway was part of the Danish state from the late 14th century until the early 19th century, and Danish became the official language of Norway roughly in the 16th century. During the Norwegian national revival in the 19th century, Bokmål arose by a gradual ‘Norwegianization’ of Danish, drawing upon local speech patterns and dialects. The other, much less used, official standard of present-day Norway is Nynorsk, which was also created in the 19th century, but this time exclusively on the basis of Norwegian rural dialects.
[3] In fact, Bokmål is only a written standard, so it cannot be said to have any phonology; however, there is a spoken variety or cluster of spoken varieties that correspond to a large extent to written Bokmål.
[4] – For Norwegian, some less familiar works within Optimality Theory include Lorentz 1996, Lunden 2007 and Johnsen 2008. Unfortunately, they can only be addressed very briefly here. This may partly be justified by the fact that while each of them, while significant and insightful, relies crucially on a rather uncommon or problematic assumption. Thus, Lorentz (1996) employs an unusual constraint banning neutralization as such and predicts secondary stress in end-stressed words consisting of two heavy syllables such as vul‘kan ‘volcano’, yet such stress is not observed in practice. Lunden (2007) assumes a constraint against stressing final open vowels on the basis that vowel length is not contrastive word-finally, yet both intuition and her own acoustic measurements suggest that words such as me‘ny ‘menu’ do have a long vowel in the stressed syllable. Johnsen’s (2008) analysis of Norwegian apparently posits a „count system“ in the sense of van der Hulst (1999b), with a constraint hierarchy that predicts the exact place of main stress (at the right edge of the word) to depend on whether the word has an even or odd number of syllables – thus, words like autodi‘dakt och amino‘plast should receive penultimate stress in contrast to kata‘rakt och kloro‘plast; this is in fact not the case.
[5] – The specific tonal contour varies between different regional standards – in modern Copenhagen speech, it typically involves low pitch on the stressed syllable and a subsequent post-tonic rise. This would be described as a L*H pitch accent in autosegmental-metrical theory – so Pierrehumbert 1980:116, although see Gussenhoven 2004:225 for a different interpretation, involving a delayed tonal target.
[6]Admittedly, some theorists have sought to describe the difference between primary and secondary stress in English exclusively in terms of presence vs absence of pitch accent, but there are empirical arguments against this view (see Ladd 1996:46-51 and Gussenhoven 2004:19-22 for references and discussion).
[7] That should then be viewed in light of the widely held opinion that Danish intonation used to be constrained by a lexical tonal distinction such as the one still found in neighbouring Swedish and Norwegian.
[8] I will be mostly using phonetic transcription in the following. Phonemic transcription may have seemed a more natural choice for the present purposes, but the issue of what is phonemic and what is phonetic as far as segments are concerned can often be rather contentious, irrelevant for this paper or both. More specifically, I will generally adhere to the ‘normalized semi-narrow’ phonetic transcription employed by Basbøll (2005:26) for Danish and the broad phonetic transcription used by Kristoffersen (2000:11) for Norwegian. Note that the Danish values of the IPA symbols in particular are often markedly different from the standard ones; for example, most front vowels and several back vowels are significantly higher than the normalized transcription would suggest.
[9]There are some differences between the analyses of the two languages in this respect. First, Basbøll considers Danish schwa a separate phoneme (2005:52), while Kristoffersen views Norwegian schwa as a posttonic allophone of /e/ (2000:21). Second, Basbøll regards Danish syllabic sonorants as expressing a lexical (≈ underlying) nuclear schwa that precedes the sonorant (2005:57-60), while Kristoffersen does not posit an underlying vowel in these cases and believes that Norwegian syllabic sonorants mostly result automatically from the normal syllabification algorithm and in the other cases are underlyingly syllabic (2000:57-59, 215-219). Examples of surface forms: Danish cykel [ˈsyɡ̊l̩], Norwegian sykkel [1sʏkːl̩] (‘bicycle’), Danish måne [ˈmɔːn̩̩] (‘moon’) (but Norwegian måne [ˈmoːn̩ə], not [2moːn̩̩]!). These differences have certain limited implications for the mechanism of stress assignment. In words like måne, the first analysis claims that the posited inherent unstressability of a phonemic lexical schwa influences the place of stress, while the second one asserts that the schwa quality itself is the allophonic result of the stress placement. In words like cykel / sykkel, the first analysis again appeals to the unstressability of schwa, while the second one presumably uses the unstressability of a syllabic sonorant instead.
[10] In Urban East Norwegian, these melodies can be expressed as L* and H*L respectively, both followed by a H– tone that spreads to the end of the Accent Phrase.
[11]This may need a clarification. Norwegian tonal accent is commonly desribed as a property of the word as a whole, because even words containing several stressed syllables may get only one accent, and the realization requires at least one post-stress syllable. Danish stød is described as a property of the syllable, because it may occur in several syllables within the same word, provided that each has some degree of stress (we accept Basbøll’s analysis here), and is realized within the syllable.
[12] The term ‘simplex words’ refers here to word forms that are not formed by affixation or compounding (another way to use the term would be in reference to all non-compound words, as in Riad 2003a). Obviously, there may be borderline cases such as final -e in dan. gade, nor. gate ‘street’; the final vowel in corresponding Swedish forms such as gata is analysed by Riad (2003a:18, 2008) as a suffix -/a/ attached to the root /gat/-. I shall abstract from this issue, as do the authors cited, since it has no significant implications for the subject at hand.
[13]Actually, the native core of simplex words are either monosyllabic or disyllabic with a schwa vowel in the second syllable (e.g. gade /ˈɡ̊aːdə/, [ˈɡ̊æːð̩] or [ˈɡ̊æːðə], ‘street’), so stress is entirely predictable based on the unstressability of schwa.
[14]Originally, the entire prosodic word would be parsed into feet and then one of these would receive primary stress, while the rest were believed to cause a rhythmic alternation of secondary stresses. Increasingly, the latter form of rhythmic footing is instead achieved in a separate procedure, subsequent to primary stress assignment (see e.g. van der Hulst 1999b:72, likewise in Roca and Johnson 1999), and for some languages, it is not done at all. Specifically, neither Danish nor Norwegian seem to show clear evidence of iterative footing, so the construction of a single foot is sufficient for most purposes.
[15]The superscript 1 and 2 in the transcription of Norwegian refer to the two tonal accents. In this context, the tonal contrast is irrelevant and both merely imply metrical stress.
[16]Kristoffersen (2000) does not see consonant length word-finally as being part of either the phonological or the phonetic output; Rice (2006) does.
[17]Kristoffersen’s conclusion is based on the study of a corpus of monomorphemic lexemes (p.149-155) and is convincing.
[18]The whole account could, however, be salvaged within a version of so-called Stratal Optimality Theory, which does recognize several levels with different constraint rankings. The grammar would then assign stress per Weight-to-Stress on one level and then cause the lengthening per Stress-to-Weight on a later level.
[19] An early version of this re-interpretation of final consonant extrametricality was already proposed in Giegerich 1985:48. It should be noted that in this case, final syllable extrametricality must be an unrelated phenomenon.
[20] Compounds will be treated in Part 2, but one may point out already here that the stress pattern of Norwegian and Swedish compounds, both „formal“ and „true“, is generally explained by their prosodic structure, which in turn is determined by their morphology or lexical specification. They are composed of not one but several prosodic domains (e.g. prosodic words), each of these domains has its own head, and that head receives stress and length. The strength relationship between the domains is determined by a parameter/constraint which is completely separate from the one applying to main stress in simplex words. It may be interesting to consider the possibility that „real“ systems with conflicting directionality such as Selkup and Chuvash can be explained in the same way.
[21] Non-sonorant consonants in Danish are all voiceless.
[22]Another solution may be inferred from the work of Gordon (1999, 2002) and others, who actually distinguish between weight for the purposes of stress and weight for the purposes of tone (both kinds of system may occur independently in the same language). Gordon’s surveys show that among consonants, it is common that only sonorants contribute weight for tone, while the same restriction is very rare for stress. It has been argued that stød is, underlyingly, a kind of tonal configuration (Riad 2000), so one can conclude that the two types of weight are both present in Danish. Here, however, the concepts of weight and moraicity will be reserved for „weight for the purposes of stress“, and it will be assumed that „weight for the purposes of tone“, including „stød weight“, is simply a set of constraints determining which morae (i.e. morae associated with which type of segments) can be tone-bearing.
[23]An exception is the fact that two identical consonants on both sides of a morpheme boundary may result in a phonetic geminate (Basbøll 2005:82).
Съпоставително изследване на мястото на ударението
в датския и норвежкия език
Част І – Прости думи
Настоящата статия разглежда в съпоставителен план принципите, определящи мястото на ударението в датския и норвежкия. Като основен източник се използват описанията на датския от Ханс Басбьол (2005) и Нина Грьонум (1998) и на норвежкия от Йерт Кристофершен (2000) и Кърт Райс (2006). В първа част се разглеждат два проблема: степените на ударение (първично и вторично) и техните характеристики и мястото на ударението в простите (непроизводни) думи.
По отношение на степените на ударение се открива сходство, доколкото и в двата езика вторичното ударение се характеризира с възможността за квантитативен контраст, а първичното ударение в добавка към това се свързва и с интонационен сигнал. Същевременно се посочва необходимостта за по-ясно различаване на фонологични импликации на ударението като гореспоменатите и чисто фонетични сигнали.
Внимателен преглед на съществуващите описания на принципите за мястото на ударението в прости думи разкрива, че различията между теоретичните подходи и обобщенията на авторите „прикриват“ реалното сходство между езиците. Оказва се, че до голяма степен датският може да бъде описан с помощта на същите правила, с които и норвежкият: първичното ударение в думата най-често пада на последната сричка, ако тя е затворена, а в противен случай на предпоследната сричка. Трудно е обаче този принцип да се формализира в датския по същия начин, както в норвежкия, тъй като ударението и в двата езика е тясно свързано с квантитета, а между квантитативните им системи има значителна разлика. Разглеждат се различни възможни решения на проблема за квантитета в датския, като в крайна сметка везните се накланят в полза на признаването на дългите гласни за такива още в подлежащата (дълбинна) репрезентация.