21.12.2012 Views

Preference Laws in Change Robert Mailhammer (Arizona State ...

Preference Laws in Change Robert Mailhammer (Arizona State ...

Preference Laws in Change Robert Mailhammer (Arizona State ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

<strong>Robert</strong> <strong>Mailhammer</strong> (<strong>Arizona</strong> <strong>State</strong> University), David Restle and Theo Vennemann<br />

(University of Munich)<br />

1. Introduction<br />

Dur<strong>in</strong>g the 1980s <strong>Robert</strong> Murray and Theo Vennemann developed an approach to language<br />

change, which, build<strong>in</strong>g on theories of markedness and naturalness, aimed at elucidat<strong>in</strong>g the<br />

motivations beh<strong>in</strong>d sound changes. From the beg<strong>in</strong>n<strong>in</strong>g this theory stood <strong>in</strong> apparent contrast<br />

to traditional approaches, which were largely descriptive. The key assumptions beh<strong>in</strong>d the<br />

preference-law approach is that certa<strong>in</strong> l<strong>in</strong>guistic structures are more preferred than others<br />

and that languages strive towards improvement <strong>in</strong> l<strong>in</strong>e with these preferences. This<br />

perspective on language change has provided theoretical underp<strong>in</strong>n<strong>in</strong>gs for more recent<br />

phonological theories, most notably Optimality Theory.<br />

This chapter <strong>in</strong>troduces the concept of preference laws, giv<strong>in</strong>g an overview of its<br />

development, its most significant features and key applications, as well as a discussion of<br />

critical po<strong>in</strong>ts raised <strong>in</strong> the literature. It is structured as follows. Section 2 <strong>in</strong>troduces the<br />

theoretical foundations and essential basics of preference-law theory. In section 3 we<br />

demonstrate its application <strong>in</strong> motivat<strong>in</strong>g sound change. In sections 4 and 5 developments and<br />

controversies <strong>in</strong>volv<strong>in</strong>g the concept of preference laws are discussed. The ma<strong>in</strong> po<strong>in</strong>ts of this<br />

chapter are summarized <strong>in</strong> section 6.<br />

2. <strong>Preference</strong> laws as a theoretical concept<br />

In a nutshell, preference laws are graded statements about quality. The concept is a<br />

generalization of Jakobson‟s markedness theory. Instead of us<strong>in</strong>g a contrast<strong>in</strong>g pair of terms<br />

“marked” vs. “unmarked”, it works with a gradation “the less marked, the”, “the more<br />

preferred, the” or simply “the better, the”. In other words, preference laws make a statement<br />

about a preferred structure with respect to a certa<strong>in</strong> parameter, <strong>in</strong> the sense that “X is more<br />

preferred than Y <strong>in</strong> terms of Z. Z is a gradable property of X and Y”. The preferences<br />

themselves are based on the criteria that Jakobson outl<strong>in</strong>ed for his markedness theory, but<br />

essentially they are theorems:<br />

- CROSSLINGUISTIC DISTRIBUTION: The more languages show a certa<strong>in</strong> property the<br />

more preferred it is.<br />

- FIRST LANGUAGE ACQUISITION: The earlier and quicker a certa<strong>in</strong> feature is acquired,<br />

the more preferred it is.<br />

- APHASIA: The longer it takes for a certa<strong>in</strong> structure to become lost <strong>in</strong> aphasia, the<br />

more preferred it is.<br />

Though Jakobson‟s orig<strong>in</strong>al markedness theory was applied to sound <strong>in</strong>ventories (phoneme<br />

systems), this concept can be applied to all human culture systems and their applications.<br />

Related approaches have been used <strong>in</strong> Natural Phonology, Morphology and Syntax as well as<br />

Optimality Theory. Similarly to Jakobson, the preference laws discussed here apply to<br />

syllable structure. The fundamental assumption is that a change <strong>in</strong> a language‟s system is<br />

always a local improvement with respect to one given parameter. Every parameter permits<br />

several configurations, and the tendency towards optimization is expressed <strong>in</strong> the hypothesis<br />

that a language will, as a rule, not possess dispreferred structures with respect to a s<strong>in</strong>gle<br />

parameter without possess<strong>in</strong>g more preferred structures at the same time. This can be<br />

formulated as the Synchronic Maxim:<br />

(1) Synchronic Maxim (Vennemann 1988: 3)


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

A language system will <strong>in</strong> general not conta<strong>in</strong> a structure on a given parameter without<br />

conta<strong>in</strong><strong>in</strong>g those structures constructible with the means of the system that are more<br />

preferred <strong>in</strong> terms of the relevant preference law.<br />

It is necessary to relativize this maxim, because occasionally gaps on the graded scale of<br />

parameters occur as a result of more global changes or language contact. In some cases gaps<br />

are motivated naturally, i.e. because some sound comb<strong>in</strong>ations are phonetically impossible,<br />

but <strong>in</strong> general this hypothesis holds, because it is based on the human natural language<br />

endowment and human culture, which form the basis of all language systems.<br />

The corollary to the Synchronic Maxim is the Diachronic Maxim, which makes a<br />

statement on how language change proceeds with respect to a certa<strong>in</strong> parameter. Accord<strong>in</strong>gly,<br />

language change affects the worst structures first, and then moves up the parametrical scale<br />

towards the end of optimization.<br />

(2) Diachronic Maxim (Vennemann 1988: 2)<br />

L<strong>in</strong>guistic change on a given parameter does not affect a language structure as long as<br />

there exist structures <strong>in</strong> the language system that are less preferred <strong>in</strong> terms of the relevant<br />

preference law.<br />

It is important to keep <strong>in</strong> m<strong>in</strong>d the local limitation of the language improvement assumption.<br />

A change <strong>in</strong> syllable structure naturally takes place on the level of the syllable. Hence, a<br />

change that is not an improvement accord<strong>in</strong>g to a preference law on the level of the syllable<br />

is motivated by a preference on another parameter and not by a syllabic preference. What is<br />

more, as will become evident from the synopsis of the most important preference laws <strong>in</strong> 3.<br />

below, sometimes preference laws on a given level can be <strong>in</strong> competiton with each other. In<br />

those cases, the precedence of one law over the other can be specific to a certa<strong>in</strong> change.<br />

For <strong>in</strong>stance, Sp escuela improves the first-syllable head of ancestral L schola [sk] <strong>in</strong> l<strong>in</strong>e<br />

with the Head Law (cf. (4) below), by remov<strong>in</strong>g the appendix s- from this syllable. However,<br />

<strong>in</strong> do<strong>in</strong>g so the word becomes longer by one syllable, which is aga<strong>in</strong>st a universal preference<br />

for shorter words, and it is also <strong>in</strong> disharmony with the Head Law and the Coda Law (see (6)<br />

below) with respect to the newly-created first syllable, which is naked and closed by a fairly<br />

strong consonant. Evidently, break<strong>in</strong>g up the <strong>in</strong>itial cluster <strong>in</strong> harmony with a phonotactic<br />

constra<strong>in</strong>t aga<strong>in</strong>st [sk] <strong>in</strong> heads has priority over optimiz<strong>in</strong>g a new syllable which does not<br />

violate the phonotactic regularities of Spanish, although it clearly does not have a preferred<br />

structure.<br />

Apply<strong>in</strong>g the preference laws diachronically can not only provide the motivation for a<br />

given sound change, it can also expla<strong>in</strong> sound changes that seem to run <strong>in</strong>to opposite<br />

directions. Dispreferred syllable structures can usually be improved <strong>in</strong> more than one way<br />

(see e.g. Vennemann 1988: 50-51 for a catalogue of syllable contact changes), and a language<br />

may use two different strategies at the same time. For <strong>in</strong>stance, “Pāli has regularly elim<strong>in</strong>ated<br />

cases of word <strong>in</strong>itial clusters by means of a cluster reduction or epenthesis” (Murray 1982:<br />

176), i.e. it simplified heads <strong>in</strong> accordance with the Head Law <strong>in</strong> two different ways: Skrt.<br />

srotas „stream’ vs. Pāli sota but Skrt. sneha „affection‟ vs. Pāli s<strong>in</strong>eha.<br />

Generally, a language will show a tendency towards one solution, but sometimes two<br />

solutions appear to occur with an almost equal distribution (see e.g. Labov 2010: ch. 7 on<br />

bifurcational sound changes <strong>in</strong> North American English). The next section will illustrate the<br />

most important preference laws as formulated <strong>in</strong> Vennemann (1988), and demonstrate their<br />

significance for language change.<br />

3. <strong>Preference</strong> laws for syllable structure<br />

2


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

In the formulation of the preference laws the concept of consonantal strength as a relational<br />

measure of “deviation from unimpeded (voiced) airflow” (Vennemann 1988: 8) is pivotal. 1<br />

Inversely related to this measure is the idea of sonority, which orders speech sounds<br />

accord<strong>in</strong>g to how vowel-like they are (see Murray 1988 for a discussion of the two concepts).<br />

Several formulations with different grades of sonority can be found <strong>in</strong> the literature, see<br />

Restle & Vennemann 2001: 1312 for examples.<br />

(3) Scale of consonantal strength<br />

INCREASING CONSONANTAL STRENGTH<br />

voiceless stops<br />

voiced stops<br />

voiceless fricatives<br />

voiced fricatives<br />

nasals<br />

lateral liquids (l-sounds)<br />

central liquids (r-sounds)<br />

glides/approximants<br />

high vowels<br />

mid vowels<br />

low vowels<br />

INCREASING SONORITY<br />

Such a categorization is aimed at expla<strong>in</strong><strong>in</strong>g the relative position of sounds to the syllable<br />

nucleus. Members of sound classes towards the bottom of the scale, i.e. the more sonorous<br />

sounds, tend to be closer to the syllable nucleus than those towards the top of the scale, i.e.<br />

the more consonantal sounds. This has been recognized <strong>in</strong> the literature (see e.g. the Sonority<br />

Sequenc<strong>in</strong>g Generalization formulated <strong>in</strong> Blev<strong>in</strong>s 1995: 210).<br />

For reasons of space we will discuss only a selection of the preference laws of for syllable<br />

structure <strong>in</strong> more detail, cover<strong>in</strong>g <strong>in</strong>trasyllabic, <strong>in</strong>tersyllabic and extrasyllabic relations.<br />

3.1 Intrasyllabic relations<br />

The preference laws for syllable structure formulated <strong>in</strong> Vennemann 1988 reflect empirical<br />

studies on syllable structure (e.g. Greenberg 1978). The basic generalization these<br />

preferences are based on is that a syllable structure is the better the more monotonous it is.<br />

Monotony is def<strong>in</strong>ed as a cont<strong>in</strong>u<strong>in</strong>g decrease of consonantal strength <strong>in</strong> the body and a<br />

cont<strong>in</strong>u<strong>in</strong>g <strong>in</strong>crease of consonantal strength <strong>in</strong> the rhyme (see Vennemann 1988: 9-10 for<br />

further details). 2 These laws also express that the optimal syllable possesses a head with a<br />

s<strong>in</strong>gle, strong consonant, a nucleus with a low monophthongal vowel and an empty coda, as<br />

e.g. pa, ta and ma.<br />

1 See Restle & Vennemann 2001: 1312 for references discuss<strong>in</strong>g the phonetic foundation of sonority concepts.<br />

2 The body comprises syllable head and nucleus, the rhyme nucleus and coda. See Vennemann 1988: 5-10 for<br />

the basic term<strong>in</strong>ology on syllable structure used here.<br />

3


3.1.1 The Head Law<br />

<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

(4) Head Law<br />

A syllable head is the more preferred,<br />

(a) the closer the number of speech sounds is to one,<br />

(b) the greater the consonantal strength value of its onset, and<br />

(c) the more sharply the consonantal strength drops from the onset towards the nucleus.<br />

(a): Empty syllable heads, i.e. heads with fewer than one speech sound, are dispreferred.<br />

This can be seen from the fact that there are no languages allow<strong>in</strong>g only empty heads, but<br />

languages bann<strong>in</strong>g empty heads are not uncommon. One such example is Tolowa<br />

(Athapaskan), which permits the follow<strong>in</strong>g syllable types, CVV, CCVV, CVC, CCVC,<br />

CCVCC, CVCC and – more rarely – CV but not VV, VC and VCC (Coll<strong>in</strong>s 1989). Northern<br />

Standard German fills empty heads of stressed and of <strong>in</strong>itial syllables with a glottal stop, e.g.<br />

chaotisch „chaotic‟ [k h ɑ.ˈʔo.t h ɪʃ]. (but Chaos [ˈk h ɑ.ɔs]), Aorta „aorta‟ [ʔɑ.ˈʔɔr.t h ɑ]. But this<br />

preference of filled syllable heads is not particularly strong, as there are <strong>in</strong>deed languages<br />

with a considerable number of naked syllables, i.e. syllables with empty heads. For <strong>in</strong>stance,<br />

<strong>in</strong> Basque about half of all words beg<strong>in</strong> with a vowel. Moreover, naked syllables may be<br />

treated differently accord<strong>in</strong>g to syllable position, e.g. language may have different rules for<br />

hiatus position (see 3.2.2 below). Furthermore, there are also languages <strong>in</strong> which the<br />

canonical syllable seems to have an empty head, e.g. the Australian languages Arrernte and<br />

Kunjen (Evans 1995: 747).<br />

Conversely, syllable heads with more than one speech sound are not considered<br />

optimal either, and this is proportionate to the number of speech sounds above one. This can<br />

be <strong>in</strong>ferred from the fact that a given language (1) will have an upper limit as to the number<br />

of speech sounds permitted <strong>in</strong> the syllable heads; (2) will permit heads with n-1 speech<br />

sounds if it permits heads with n (n > 1) speech sounds, and (3) will have more limitations as<br />

the number of speech sounds <strong>in</strong> the syllable head <strong>in</strong>creases. For <strong>in</strong>stance, Standard German,<br />

permits heads filled with a maximum of three speech sounds, and hence also with two and<br />

one, conform<strong>in</strong>g to the first two postulates. That it is also <strong>in</strong> l<strong>in</strong>e with the third can be seen<br />

once one exam<strong>in</strong>es which sequence of consonants can also occur with the reverse order (this<br />

is not valid for all clusters consist<strong>in</strong>g of obstruents plus sonorants, e.g. [kr-] but *[rk-], [ʃm]<br />

but *[mʃ]), and if it is borne <strong>in</strong> m<strong>in</strong>d that the roughly 20 permitted speech sounds do not form<br />

20*20*20, i.e. 8000, triple consonant clusters, but only four, namely [ʃpr-], [ʃpl-], [ʃtr-] and<br />

[skr-] (spr<strong>in</strong>gen „jump‟, Splitter „chip‟, streng „strict‟ and Skrupel „scruple‟), possibly six if<br />

[skl-] and [str-], occurr<strong>in</strong>g <strong>in</strong> little-<strong>in</strong>tegrated loanwords (e.g. Sklerose „sclerosis‟ and Strip<br />

„strip‟), are added. 3<br />

In language change one observable effect of this preference law is the frequent<br />

shorten<strong>in</strong>g of syllable heads. For <strong>in</strong>stance, heads <strong>in</strong> Old Indic (Vedic and Sanskrit) could<br />

comprise up to three speech sounds, whereas <strong>in</strong> Middle Indic (e.g. Pāli) consonant clusters<br />

were no longer permitted <strong>in</strong> syllable heads at all. A development <strong>in</strong> the opposite direction, i.e.<br />

a systematic lengthen<strong>in</strong>g of syllable heads does not seem to occur. Lengthen<strong>in</strong>g of syllable<br />

heads usually has a specific reason that has noth<strong>in</strong>g to do with syllable structure. The<br />

epenthetic change of sr > str <strong>in</strong> Pre-Proto-Germanic is a more global change that elim<strong>in</strong>ates<br />

the sequence sr generally and not only <strong>in</strong> syllable heads; the lengthen<strong>in</strong>g of syllable heads is<br />

just a side-effect: Gmc. + strauma- „river, stream‟ vs. OI srávati „(3sg) flows‟, but also PGmc.<br />

3 Whether [pfr-], [pfl-] and [tsv-] have to be <strong>in</strong>cluded, depends on whether the affricates [pf]<br />

and [ts] are counted as monophonematic or diphonematic.<br />

4


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

+ swester- < PIE + swesr- with epenthentic t <strong>in</strong> the syllable coda or the syllable contact<br />

respectively. 4 Complex syllable heads can also result from syncope, but this is evidently not<br />

aimed at improv<strong>in</strong>g syllable structure but at shorten<strong>in</strong>g the word.<br />

(b): Many languages do not allow the least consonantal sounds to form syllable heads,<br />

e.g. all vowels <strong>in</strong> Standard German. This tendency towards maximally consonantal heads is<br />

especially apparent <strong>in</strong> word-<strong>in</strong>itial position, as consonants tend to weaken by assimilation<br />

word-medially. Spanish, for <strong>in</strong>stance, tolerates both /ɾ/ and /r/ as word-medial syllable heads<br />

but only /r/ <strong>in</strong> word-<strong>in</strong>itial heads: pero „but‟, perro „dog‟, but only rojo [r-] „red‟. Many<br />

languages do not permit voiced fricatives word-<strong>in</strong>itially but only voiceless ones (e.g. Old<br />

English), others aspirate voiceless stops <strong>in</strong> this position (e.g. Modern English).<br />

(c): This preference can be illustrated with several well-known cases. Standard<br />

German allows strong consonants (obstruents) before sonorants but not weak consonants<br />

(sonorants), not even if they are relatively stronger: Knie „knee‟, Platz, „place‟, blank „bare‟,<br />

groß „big‟, frei „free‟, schmal „narrow‟, Schnee „snow‟, schlimm „bad‟, schräg „diagonal‟, but<br />

*mn-, *ml-, *mr-, *nl-, *nr-, *lr-. The history of the Germanic languages shows that [h] too is<br />

a weak consonant. It derived from a Proto-Germanic velar fricative [x], which <strong>in</strong> turn goes<br />

back to a PIE velar stop [k]. In all Germanic languages except Icelandic, [h] is then lost <strong>in</strong><br />

syllable heads before consonants, and <strong>in</strong> many varieties of English /h/ is already be<strong>in</strong>g lost<br />

before vowels (e.g. it has been lost <strong>in</strong> Cockney English), whereas it has been preserved<br />

before /w/ <strong>in</strong> others (e.g. <strong>in</strong> varieties of American English). In Icelandic, /hw/ is threatened by<br />

a more general process of strengthen<strong>in</strong>g, [w] > [v]; as far as [w] does not resist this<br />

development <strong>in</strong> this cluster, [h] is strengthened <strong>in</strong>stead to ma<strong>in</strong>ta<strong>in</strong> a sufficient fall <strong>in</strong><br />

consonantal strength: /hw/ > /hv-/ > /kv-/, e.g. <strong>in</strong> hvítur /hwitʏr/ or /kvitʏr/ „white‟.<br />

The well-known fate of word-<strong>in</strong>itial k- <strong>in</strong> English is <strong>in</strong>structive at this po<strong>in</strong>t as well.<br />

Like Modern German, Old English possessed the clusters kn-, kl-, kr-, kw-, kV-. In Late<br />

Middle English the worst of these (kn-) became unstable, and [k] was subsequently lost, cf.<br />

ModE knee. Of the rema<strong>in</strong><strong>in</strong>g groups kl- represents the worst structure accord<strong>in</strong>g to the<br />

theory and is already under attack <strong>in</strong> certa<strong>in</strong> varieties, as predicted by the Diachronic Maxim<br />

<strong>in</strong> (2) above. By contrast, the groups to the right-hand end of the scale – those with the<br />

weaker slope consonants – are completely stable <strong>in</strong> all varieties: kr- and kw- can only be<br />

elim<strong>in</strong>ated after kl- has disappeared, and it is predicted that kr- becomes unstable before kw-,<br />

unless [w] would strengthen to [v] like <strong>in</strong> Islandic and German (see Lutz 1991: ch. IV.B.2 for<br />

the developments <strong>in</strong> English and ch. IV.B.3 for the developments <strong>in</strong> the other Germanic<br />

languages<br />

However, the Romance languages illustrate another way to get rid of word-<strong>in</strong>itial<br />

consonant clusters, namely by weaken<strong>in</strong>g the second consonant rather than strengthen<strong>in</strong>g or<br />

delet<strong>in</strong>g the first. This can be shown with what happened to <strong>in</strong>herited clusters consist<strong>in</strong>g of<br />

obstruent plus liquid, after obstruent plus nasal had been elim<strong>in</strong>ated already <strong>in</strong> Lat<strong>in</strong>.<br />

In many Romance languages the structurally worst structures became subsequently<br />

unstable, namely those consist<strong>in</strong>g of obstruent plus l. In a first step l developed <strong>in</strong>to a palatal<br />

glide after voiceless obstruents, and then the whole group gradually ended up as a voiceless<br />

alveo-palatal fricative ([ʃ], spelt ch): Lat<strong>in</strong> plumbum „lead‟, clavis „key‟, flamma „flame‟ > P<br />

chumbo, chave, chama. In a second step, l <strong>in</strong> new loans changed to r follow<strong>in</strong>g a voiced stop<br />

and follow<strong>in</strong>g a voiceless obstruent <strong>in</strong> new loans, compare Spanish blanco „white‟, obligar<br />

„pledge‟, regla „rule‟, plancha „board‟, clavo „nail‟, flota „fleet‟ with Portugese branco,<br />

obrigar, regra, prancha, cravo, frota. The aim of these developments was an improved<br />

4 To avoid ambiguity, the raised cross „ + ‟ <strong>in</strong>dicates a reconstructed form, the asterisk „*‟ marks an<br />

ungrammatical form.<br />

5


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

system, which is be<strong>in</strong>g underm<strong>in</strong>ed by more recent loanwords – sometimes words have been<br />

borrowed twice, e.g. flauta ~ frauta „flute‟.<br />

(5) Head improvement <strong>in</strong> Portuguese<br />

*KN- *Kl- Kr-<br />

<strong>in</strong>creas<strong>in</strong>g quality of syllable head<br />

Heads consist<strong>in</strong>g of obstruent plus r have so far been stable everywhere, they even represent<br />

partly the target structure <strong>in</strong> the elim<strong>in</strong>ation process of l-clusters (cf. Vennemann 1989: 17-21<br />

[1933: 326-330] on head clusters)<br />

3.1.2 The Coda Law<br />

(6) Coda Law<br />

A syllable coda is the more preferred,<br />

(a) the smaller the number of speech sounds,<br />

(b) the lower the consonantal strength of its offset, and<br />

(c) the more sharply the consonantal strength drops from the offset towards the<br />

consonantal strength of the preced<strong>in</strong>g syllable nucleus.<br />

(a) and (b): Sanskrit progressively reduces word-f<strong>in</strong>al codas from outside to <strong>in</strong>side until<br />

maximally the <strong>in</strong>nermost consonant rema<strong>in</strong>s: adan „eat<strong>in</strong>g‟, replac<strong>in</strong>g ad-ant-s (cf. L edēns <<br />

ed-ent-s, acc. edentem). The only exceptions are clusters consist<strong>in</strong>g of /r/ plus stop, i.e.<br />

exactly those clusters with the steepest possible drop <strong>in</strong> consonantal strength: āvart „(3sg)<br />

turned‟, amārṭ „(3sg.) wiped‟, vark „(3sg.) bent‟.<br />

(b): After the implementation of Kl<strong>in</strong>genheben‟s Law, accord<strong>in</strong>g to which all l<strong>in</strong>gual coda<br />

consonants changed to /r/ and all labial and velar consonants to /w/, Hausa permitted only<br />

these two weakest consonants <strong>in</strong> the coda: + ma.za.ma.za „very fast‟ > + maz.ma.za ><br />

mar.ma.za; + ma.kaf.ni.ya „a bl<strong>in</strong>d female‟ > ma.kaw.ni.ya (~ ma.kā.fo „a bl<strong>in</strong>d male‟); + hag.ni<br />

„left side‟ > haw.ni (~ ba.ha.go „a left-handed person), etc. (cf. Kl<strong>in</strong>genheben 1928).<br />

3.1.3 The Nucleus Law<br />

(7) Nucleus Law<br />

A syllable nucleus is the more preferred,<br />

(a) the closer the number of its speech sounds is to one,<br />

(b) the lower the consonantal strength of its speech sounds.<br />

(a): If a language has nuclei of a certa<strong>in</strong> length (> 1), then it always has also shorter nuclei<br />

(≥ 1), and the upper limit is low <strong>in</strong> all languages (≤ 3). All languages have monophthongs but<br />

not necessarily diphthongs, and triphthongs are rare but do occur, e.g. <strong>in</strong> Portuguese.<br />

Diachronically, this preference surfaces <strong>in</strong> the shape of crossl<strong>in</strong>guistically very frequent<br />

spontaneous monophthongizations. Diphthongizations, by contrast, never occur<br />

spontaneously, they are always motivated by some other factors with<strong>in</strong> a given system.<br />

(b): Low vowels may function only as nuclei and as offglides of diphthongs (though they<br />

do not make good offglides). The higher a vowel, the more it tends to be marg<strong>in</strong>alized,<br />

especially if adjacent to sounds that can also function as nuclei. Consonants generally are<br />

poor nuclei, especially if they are strong. This becomes clear from the fact that only few<br />

languages have consonantal nuclei, and if they do, these tend to occupy cont<strong>in</strong>uous sections<br />

6


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

towards the sonorous end on the scale of consonantal strength, e.g. only r (like <strong>in</strong> Croatian,<br />

e.g. Krk, an island <strong>in</strong> the Adriatic Sea) or r and l (like <strong>in</strong> Sanskrit), or these and the nasals<br />

(like <strong>in</strong> English unstressed syllables, e.g. hammer, saddle, mutton, bottom).<br />

(8) Corollary to the Nucleus Law<br />

The shell of a syllable is the more subject to severe restrictions, the higher the<br />

consonantal strength of its nucleus.<br />

This corollary is expla<strong>in</strong>ed by the fact that both the Head and Coda Law cannot easily be<br />

fully observed if the nucleus is filled with a very consonantal speech sound. If, for <strong>in</strong>stance, a<br />

speech gesture <strong>in</strong> German conta<strong>in</strong>s /s/ as its nucleus, this means that <strong>in</strong> order to keep the shell<br />

monotonous it is restricted to stops, pst „hush‟. This also expla<strong>in</strong>s the only exception to the<br />

rule for syllable heads <strong>in</strong> Tolowa (cf. 3.1.1 above), which is that a syllable with /n/ as nucleus<br />

does not permit any speech sound <strong>in</strong> its shell.<br />

3.2 Intersyllabic Relations<br />

Languages can differ not only with respect to their syllable structure, but also with respect to<br />

how they str<strong>in</strong>g syllables together. This has been the subject of a fruitful direction of<br />

phonological research. Perhaps the most well-known generalization is the pr<strong>in</strong>ciple of Head<br />

Maximization (maximal onset pr<strong>in</strong>ciple, CV-rule, onset-first pr<strong>in</strong>ciple, left-precedence<br />

pr<strong>in</strong>ciple, ONSET, e.g. already Varma 1929, Allen 1953, Bell 1977, Selkirk 1982, Clements<br />

1990, Pr<strong>in</strong>ce & Smolensky 2002). This pr<strong>in</strong>ciple, however, at least as a universal rule of<br />

language structure, is <strong>in</strong>accurate. This can easily be seen from the fact that different<br />

languages – even one and the same language at different chronological stages – can syllabify<br />

the same sequence of sounds differently.<br />

This has been well-known s<strong>in</strong>ce antiquity, as clusters of obstruent (K) and sonorant<br />

(R) (muta cum liquida) <strong>in</strong> <strong>in</strong>tervocalic position caused problems for Greek and Lat<strong>in</strong> poets if<br />

the first vowel was short: the syllabification V.KRV meant that the first syllable was short,<br />

and hence light, whereas the syllabification VK.RV meant that the first syllable was long, and<br />

thus heavy. In the metric system of a quantity language this is a significant difference. As a<br />

result, syllabification was used to manipulate syllable weight. Antique poets even co<strong>in</strong>ed a<br />

term for mak<strong>in</strong>g a syllable with short nucleus light by tautosyllabification of a muta cum<br />

liquida cluster, cf. correptio Attica <strong>in</strong> Classical Greek. But varieties of antique languages<br />

without correptio, <strong>in</strong> which KR cluster are heterosyllabified, have word-<strong>in</strong>itial KR-clusters<br />

too, cf. kr- <strong>in</strong> older Greek krókos „crocus‟, but pikrós „spicy‟ with k $ r (Allen 1974: 101-102).<br />

The pr<strong>in</strong>ciple of Head Maximization would require the syllabification [pi.krós]. Moreover,<br />

the opposite of correptio – namely productio (lengthen<strong>in</strong>g) <strong>in</strong> Late Lat<strong>in</strong>, compare Classical<br />

Lat<strong>in</strong> <strong>in</strong>tegrum [ˈ<strong>in</strong>.te.grum] but Late Lat<strong>in</strong> [<strong>in</strong>.ˈteg.ro] - is also well-known, which shows<br />

that head maximization cannot even be considered a universal tendency.<br />

S<strong>in</strong>ce the discussion of this and similar problems <strong>in</strong> Vennemann 1972, Murray &<br />

Vennemann 1982, 1983, Lutz 1985, 1986, Clements 1990, Vennemann 1987, 1988a: 40-55 –<br />

s<strong>in</strong>ce 1982 known as “syllable contact” (which refers to the connection between two speech<br />

sounds A and B at the syllable boundary A $ B) – it has become clear that syllabification<br />

depends strongly on the relative consonantal strength of the speech sounds <strong>in</strong>volved. In<br />

Faroese, for <strong>in</strong>stance, muta cum liquida clusters (except tl) always belong to the second<br />

syllable, but not <strong>in</strong> closely related Icelandic, which tautosyllabifies only fortis obstruents and<br />

r but not l. As a matter of fact, Icelandic always syllabifies V $ MLV <strong>in</strong> the case of fortis M<br />

and L = r (and for L = v and = j, which, as former glides, are even weaker than r), but<br />

VM $ LV <strong>in</strong> the case of L = l and for all stronger consonants as well as for lenis M (as well as<br />

for all weaker onset consonants). Tautosyllabification can unambiguously be identified from<br />

7


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

the effect of open syllable lengthen<strong>in</strong>g, and conversely, heterosyllabification from shorten<strong>in</strong>g.<br />

This new allocation of quantity affects all old short and long vowels as well as diphthongs<br />

(cf. Vennemann 1972, 1978).<br />

Languages can also differ with respect to the connection between syllabification and<br />

prosodic properties, e.g. stress or the position <strong>in</strong> the word. This section covers the two ma<strong>in</strong><br />

laws apply<strong>in</strong>g to <strong>in</strong>tersyllabic relations, the Syllable Contact Law and the Hiatus Law. For<br />

additional laws, namely the Law of Initials and the Law of F<strong>in</strong>als see Vennemann 1988 and<br />

Restle & Vennemann 2001.<br />

3.2.1 The Syllable Contact Law<br />

(9) Syllable Contact Law<br />

A syllable contact is the better the greater the strength difference between its second and<br />

first speech sound.<br />

This law was already illustrated <strong>in</strong> 3.2 above. Tautosyllabify<strong>in</strong>g V $ MLV <strong>in</strong> Icelandic, as<br />

mentioned above, avoids a contact M $ L exactly for the twelve least preferred cases, i.e. for<br />

those <strong>in</strong> which the above-mentioned difference <strong>in</strong> strength is smallest, <strong>in</strong> this case even<br />

negative. This becomes especially clear if the consonants are arrayed on a numerically<br />

<strong>in</strong>terpreted scale of strength, as <strong>in</strong> (10), cf. Vennemann 1972: 6.<br />

(10) 1 2 3 4 5 6<br />

v r l m d t<br />

j n b p<br />

g k<br />

f s<br />

This permits the statement that two consonants only form a syllable contact A $ B <strong>in</strong> Icelandic<br />

if the critical difference <strong>in</strong> consonantal strength, KS(B) – KS(A), is equal to or greater than -<br />

3. For <strong>in</strong>stance, this is the case with /pl/: KS(/l/) - KS(/p/) = -3; hence epli „apple‟ is<br />

syllabified /ep.li/, which can be seen from the short stressed vowel and the preaspirated fortis<br />

stop [ˈe h p.li]. But it is not the case with /pr/: KS(/r/) - KS(/p/) = -4; therefore skopkra „to roll‟<br />

is syllabified /sko.pra/, recognizable from the long stressed vowel and the postaspirated fortis<br />

stop, [skɔ:.p h ra]. Like the other preference laws, the contact law manifests itself <strong>in</strong> sound<br />

changes which improve syllable contacts (see Vennemann 1988: 50-55 for a list of relevant<br />

changes).<br />

3.2.2 The Hiatus Law<br />

(11) Hiatus Law<br />

A hiatus is dispreferred, <strong>in</strong> particular the more properties the two vowels <strong>in</strong>volved<br />

share, and also the closer one of them is phonetically to a glide.<br />

From part (a) of the Head Law (see (4) above), which disfavours naked syllables, it can be<br />

<strong>in</strong>ferred that a hiatus is not a preferred structure. This can also be seen from the fact that<br />

many languages do not allow them at all (accord<strong>in</strong>g to Bell & Hooper 1978: 8 this is valid for<br />

about half of the world‟s languages). To elim<strong>in</strong>ate hiatus structures languages use<br />

mechanisms that are similar to those that elim<strong>in</strong>ate empty syllable heads as well as some<br />

KS<br />

8


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

additional ones, such as the <strong>in</strong>sertion of a glide and others (see Restle & Vennemann 2001:<br />

1318-1319 for further details).<br />

3.3 Extrasyllabic relations<br />

Languages also exhibit preferences with respect to the relationship between syllables and<br />

stress as well as rhythm. This section highlights some key pr<strong>in</strong>ciples.<br />

3.3.1 Syllabification and stress<br />

Although onset maximization as a simple pr<strong>in</strong>ciple is <strong>in</strong>correct, it is possible to state<br />

conditions for a relatively good syllabification. These are graded and <strong>in</strong>volve multiple factors,<br />

which first became apparent from research <strong>in</strong> historical phonology <strong>in</strong>vestigat<strong>in</strong>g the way<br />

words are separated <strong>in</strong> handwritten manuscripts and how this is sensitive to syllabification<br />

(Lutz 1985, 1986). This section briefly summarizes the results of this work (see Restle &<br />

Vennemann 2001: 1320-1321 for further details).<br />

(12) General Law of Syllabification (Lutz‟s Law)<br />

The syllabification of two syllables is the more preferred<br />

(a) the better the result<strong>in</strong>g syllable contact, and<br />

(b) the better this syllable contact is embedded.<br />

Part (a) of the law refers to the preferences stated by the Contact Law (cf. 3.2.1 above). In<br />

order to understand part (b) it is necessary to def<strong>in</strong>e the embedd<strong>in</strong>g of a syllable as well as<br />

other related terms and to give relevant preference laws.<br />

(13) Syllable Contact Bed<br />

Let A $ B be the syllable contact of a sequence of syllables S1.S2. The onset syllable of<br />

the contact is def<strong>in</strong>ed as the rhyme of S1 m<strong>in</strong>us A, i.e. m<strong>in</strong>us its offset, and the offset<br />

syllable as the body of S2 m<strong>in</strong>us B, i.e. its onset. The sequence of a contact‟s onset<br />

syllable and its offset syllable is called syllable contact bed.<br />

(14) Syllable Embedd<strong>in</strong>g Law<br />

The bed of a syllable contact is the better the greater the rhyme attractiveness of its<br />

onset syllable and the greater the body attractiveness of its offset syllable.<br />

(15) Rhyme Attractiveness<br />

This is def<strong>in</strong>ed as how well a syllable is capable to attract the onset of the follow<strong>in</strong>g<br />

syllable as part of its rhyme.<br />

(16) Body Attractiveness<br />

This is def<strong>in</strong>ed as how well a speech sequence is capable to attract the offset of a<br />

preced<strong>in</strong>g syllable as part of its body.<br />

(17) Law of Rhyme Attractiveness<br />

The rhyme attractiveness of a syllable (i.e. heterosyllabification <strong>in</strong>to it) is the greater<br />

(a) the more it is stressed <strong>in</strong> relation to the follow<strong>in</strong>g syllable,<br />

(b) the shorter its rhyme, and<br />

(c) the lower the consonantal strength of its rhyme, <strong>in</strong> particular its offset.<br />

(18) Law of Body Attractiveness<br />

The body attractiveness of a syllable (i.e. heterosyllabification <strong>in</strong>to it) is the greater<br />

9


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

(a) the more it is stressed <strong>in</strong> relation to the preced<strong>in</strong>g syllable,<br />

(b) the shorter its body is, and<br />

(c) the lower the consonantal strength of its body, <strong>in</strong> particular its onset.<br />

3.3.2 Syllable structure and stress<br />

Syllable structure and stress are <strong>in</strong>terrelated <strong>in</strong> multiple ways, but all seem to conform to an<br />

overarch<strong>in</strong>g preference law (see Vennemann 1988: 58 for further details):<br />

(19) Law of Stressed Syllables<br />

Complexities of syllable structure are the less favoured the less rhythmic prom<strong>in</strong>ence<br />

the syllable possesses.<br />

The clearest illustrations of this law are constra<strong>in</strong>ts on syllable structure that are categorically<br />

tied to word stress. For <strong>in</strong>stance, unstressed syllables <strong>in</strong> Icelandic can only have [ɪ], [ʏ] and<br />

[a] as nuclei, whereas stressed syllables have a much wider variety of phonologically<br />

contrastive nuclei.<br />

3.3.2 Syllable structure and rhythm<br />

There are also preferences govern<strong>in</strong>g the relationship between syllable structure and<br />

rhythmical types as well as with syllable weight, which is also relevant to rhythm.<br />

(20) Rhythm Law<br />

Complexities of syllable structure are the less favoured the smaller the rhythmic unit<br />

is.<br />

For <strong>in</strong>stance, English as a stress timed language possesses a complex syllable structure,<br />

whereas a syllable timed language such as Spanish tends to have a less complex structure,<br />

and mora timed languages seem to prefer fairly simple syllable structures (see Dufter 2003<br />

for a detailed discussion of rhythmical types). There is also a well-known connection between<br />

dynamic stress and moraicity, which has also been called Prokosch‟s Law (see Vennemann<br />

1988: 30).<br />

(21) Weight Law<br />

In stress accent languages an accented syllable is the more preferred the closer its<br />

syllable weight is to two moras, and an unaccented syllable is the more preferred the<br />

closer its weight is to one mora.<br />

In other words, stressed syllables are preferably bimoric and unstressed syllables preferably<br />

unimoric. This law has often been used to motivate quantity changes, such as lengthen<strong>in</strong>g of<br />

open monosyllables <strong>in</strong> some Germanic dialects, such as Old English (compare OE bī, nū, swā<br />

vs. Goth. bi, nu, swa).<br />

4. Developments <strong>in</strong> the concept of preference laws<br />

As mentioned <strong>in</strong> 2. above, the theory of l<strong>in</strong>guistic preferences can be seen as a generalization<br />

of Jakobson‟s markedness concept. Historically, the direct forerunners are Natural Generative<br />

Grammar and Natural Generative Phonology (Hooper 1976). Vennemann (1972) discusses<br />

problems of syllabification <strong>in</strong> Icelandic, consider<strong>in</strong>g universal <strong>in</strong>terrelationships between<br />

syllable structure and consonantal strength, which also manifest themselves <strong>in</strong> probabilistic<br />

statements about language change. Although Hooper (1976) is not primarily concerned with<br />

diachrony, she discusses implications for pr<strong>in</strong>ciples of diachronic change <strong>in</strong> connection with<br />

10


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

her pr<strong>in</strong>ciples of optimal syllable structure which represent targets <strong>in</strong> processes of language<br />

change.<br />

However, a concrete theory based on l<strong>in</strong>guistic preferences and its application to<br />

language change was proposed <strong>in</strong> Vennemann 1983a and b, as a type of theory that can<br />

expla<strong>in</strong> the “regularities with<strong>in</strong> a certa<strong>in</strong> doma<strong>in</strong> by turn<strong>in</strong>g to theories that are not theories<br />

for that particular doma<strong>in</strong>” (Vennemann 1983b: 9), as opposed to generative theories that<br />

attempt to expla<strong>in</strong> regularities by virtue of deductive <strong>in</strong>ferences with<strong>in</strong> the same doma<strong>in</strong>, <strong>in</strong><br />

this case, grammar. Consequently, the preferences themselves are motivated by<br />

extral<strong>in</strong>guistic theories, such as “phonetic theories, theories of learn<strong>in</strong>g, semiotic theories,<br />

theories of communication, etc.” (Vennemann 1983b: 13). Such a theory makes statements<br />

about a space Q <strong>in</strong> which all real languages are situated and about what is possible and what<br />

is impossible for languages <strong>in</strong> Q. As languages are not evenly distributed with<strong>in</strong> Q, one<br />

arrives at a theory operat<strong>in</strong>g on the basis of this distribution call<strong>in</strong>g it preference, markedness,<br />

naturalness etc. As Vennemann (1983b: 11) po<strong>in</strong>ts out, it is exactly this “concept of rank<br />

order” that sets preference theories apart from purely descriptive theories, and that expla<strong>in</strong>s<br />

l<strong>in</strong>guistic structures, not <strong>in</strong> a deductive sense, but <strong>in</strong> a sense that Vennemann (1983b: 13)<br />

calls “elucidation”, i.e. by shedd<strong>in</strong>g light on a particular structure or development, which can<br />

often be taken as an explanation. Discuss<strong>in</strong>g the issue of expla<strong>in</strong><strong>in</strong>g language change,<br />

Vennemann (1983b: 18f) also po<strong>in</strong>ts out that a theory of preferences <strong>in</strong> connection with a list<br />

of possible changes, a “closed catalogue”, can actually be falsified, thus be<strong>in</strong>g capable of<br />

deductive explanations. Such a catalogue of changes on a macro-level is found <strong>in</strong> Bartsch &<br />

Vennemann 1982: 153, and a nearly “complete” list of syllable-contact improv<strong>in</strong>g sound<br />

changes was assembled <strong>in</strong> Vennemann 1988: 50f.<br />

Further developments took place <strong>in</strong> subsequent work by Vennemann, Murray and Lutz <strong>in</strong><br />

discover<strong>in</strong>g and substantiat<strong>in</strong>g preference laws for syllable structure (see Murray 1982, 1988,<br />

Murray & Vennemann 1982, 1983, Vennemann 1986a, b, 1987, 1988, Lutz 1985, 1986 and<br />

the summary <strong>in</strong> Restle & Vennemann 2001). This was meant to be the beg<strong>in</strong>n<strong>in</strong>g of a new<br />

direction <strong>in</strong> research <strong>in</strong> that future energies ought to be directed at assembl<strong>in</strong>g catalogues of<br />

language change and at discover<strong>in</strong>g new preferences (Vennemann 1983b: 24f).<br />

A further development can be seen <strong>in</strong> the ways the idea of preference laws was<br />

developed from the 1990s onwards. Beg<strong>in</strong>n<strong>in</strong>g with the elaborations <strong>in</strong> Vennemann 1993,<br />

which showed the value of this concept for syntactic change, preferences have cont<strong>in</strong>ued to<br />

play a major role <strong>in</strong> Natural Phonology (see papers <strong>in</strong> Hurch & Rhodes 1996) and also<br />

Optimality Theory (see the discussions <strong>in</strong> Dziubalska-Kołaczyk 2001 and Holt 2003).<br />

Moreover, a theory of l<strong>in</strong>guistic preferences seems to be an ideal complementation to cases of<br />

diffusional and bifurcational sound changes. Beg<strong>in</strong>n<strong>in</strong>g with Labov‟s (1981) sem<strong>in</strong>al paper, it<br />

became <strong>in</strong>creas<strong>in</strong>gly apparent that Neogrammarian sound laws represent but one end of a<br />

cont<strong>in</strong>uum of sound changes, namely those with a particular thorough application (see also<br />

the discussion <strong>in</strong> Vennemann 2000a). This is underl<strong>in</strong>ed by the cases like ModE cradle and<br />

saddle and other bifurcational changes, which cannot be adequately accommodated with<strong>in</strong><br />

the Neogrammarian framework (see Murray 2005 for a discussion and Labov 2010: ch. 7 for<br />

further details on bifurcational changes). 5 In the description and elucidation of the breakdown<br />

of phonological quantity <strong>in</strong> the Germanic languages a preference-based framework (syllable<br />

cut theory) has proven to be useful, because the evaluative character of such a theory can<br />

expla<strong>in</strong> why structures change, and also <strong>in</strong> which direction they change (see e.g. Vennemann<br />

2000b, Murray 2000, <strong>Mailhammer</strong> 2009). Consequently, the notion that bifurcational changes<br />

5 OE cradol and sadol change <strong>in</strong> different directions towards Middle English despite an identical (relevant)<br />

phonological structure. The former develops smooth syllable cut on its accented syllable (with Open Syllable<br />

Lengthen<strong>in</strong>g), the latter, abrupt syllable cut (which closes the syllable ambisyllabically and therefore keeps the<br />

vowel shot), see Murray 2005 and <strong>Mailhammer</strong> 2007, 2009 for further details.<br />

11


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

are the result of an “unstable structure” be<strong>in</strong>g elim<strong>in</strong>ated <strong>in</strong> different ways (Labov 2010: 156)<br />

can be rendered mean<strong>in</strong>gful by stipulat<strong>in</strong>g why a given structure has to be seen as unstable<br />

and by predict<strong>in</strong>g the result of the change, even if the change itself cannot be predicted.<br />

Vennemann (2000) provides another especially useful application of a preference-based<br />

framework, namely <strong>in</strong> the area of etymology, connect<strong>in</strong>g words that are deemed unrelated<br />

accord<strong>in</strong>g to the Neogrammarian concept. Accord<strong>in</strong>g to this approach, words <strong>in</strong> daughter<br />

languages can be the result of differently implemented constra<strong>in</strong>ts, similarly to the<br />

cradle/saddle case <strong>in</strong> Middle English.<br />

As Vennemann (1993: 321) po<strong>in</strong>ts out, preference theory is a way out of the classical<br />

argument whether language change is fashion-like or functional. As language change is seen<br />

as language improvement, this is a “unified pr<strong>in</strong>ciple” to conceptualize language change,<br />

which accommodates “both the conscious and subconscious and both the socially and<br />

structurally motivated changes, as well as those motivated by the needs of communication”.<br />

It seems that this is the greatest achievement of this theory, which withstands criticism that<br />

has been directed at it (see the follow<strong>in</strong>g section). From the perspective of application, a great<br />

strength of preference-based approaches is the motivation of bifurcational and diffusional<br />

changes.<br />

5. Criticism of preference theory<br />

Criticism aga<strong>in</strong>st a theory of l<strong>in</strong>guistic preferences has focused ma<strong>in</strong>ly two issues, namely<br />

their empirical foundation and the notion that language change is seen as language<br />

improvement (see e.g. Berg 1990 and Hill 2009)<br />

As Vennemann (1988: 4) asserts, preference laws are theorems derived from universal<br />

properties of the human communicative capacity, and no attempt is made to justify them, as<br />

this would fall outside the doma<strong>in</strong> of l<strong>in</strong>guistics, though this is not considered an<br />

un<strong>in</strong>terest<strong>in</strong>g question (see Vennemann 1983b: 9. The need for a solid empirical foot<strong>in</strong>g<br />

stems from the perspective of laws as generalisations based on the data specify<strong>in</strong>g what<br />

happens under which circumstances. This is perhaps due to the term “laws” <strong>in</strong> the title of<br />

Vennemann (1988). As Berg (1990: 570) suggests, the term “tendencies” would have maybe<br />

been less problematic to some readers. The term “preference laws” simply refers to the<br />

qualitative evaluation of l<strong>in</strong>guistic structures. As assumptions they need not be based on<br />

quantitative empirical data, they are tested directly by the data, but not <strong>in</strong> the way generative<br />

laws are tested, s<strong>in</strong>ce we are deal<strong>in</strong>g with statements about better and worse structures and<br />

not with simple output rules. For <strong>in</strong>stance, no one would question the preferred nature of<br />

open syllables based on crossl<strong>in</strong>guistic data, even though there is no quantitative study for a<br />

sizeable body of the languages of the world and even though there may be languages <strong>in</strong><br />

which the canonical syllable is closed. Hence, the supposed lack of empirical foundation<br />

stems from the view of the preference laws as <strong>in</strong>ductively based, whereas <strong>in</strong> reality they are<br />

hypotheses.<br />

The second po<strong>in</strong>t of criticism centres on the idea that language change is always an<br />

improvement with respect to a certa<strong>in</strong> parameter (see especially Hill 2009: 244). Aga<strong>in</strong>, this<br />

is not a notion that is empirically supported; it is an axiomatic statement that is founded on<br />

the idea that human development is motivated by aim<strong>in</strong>g at improvement rather than<br />

worsen<strong>in</strong>g. The opposite or the idea that change is unmotivated seems difficult to imag<strong>in</strong>e.<br />

That there are changes for the worse is obvious but whether such a change was <strong>in</strong>tentional (as<br />

a s<strong>in</strong>gle development) is very doubtful. Consequently, there is no sensible alternative but to<br />

assume that a change lead<strong>in</strong>g to a dispreferred structure was motivated differently. This is the<br />

solution to the puzzle <strong>in</strong> Berg (1990: 570): “I can see no reason whatsoever why we should<br />

follow Vennemann <strong>in</strong> believ<strong>in</strong>g that ora > gora is a syllable structure change whereas la vita<br />

> la ita is not.” If the loss of the head consonant <strong>in</strong> the second case was not an improvement<br />

12


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

on some other level it can only have been either a random change or a deliberate move to<br />

decrease the quality of the syllable head of vita, both of which we f<strong>in</strong>d difficult to believe.<br />

The example adduced by Berg can be seen as a case of <strong>in</strong>tervocalic assimilatory weaken<strong>in</strong>g,<br />

which is then clearly not a change motivated by syllable structure.<br />

6. Summary<br />

This chapter illustrates and discusses the concept of preference laws and their application <strong>in</strong><br />

problems of historical phonology. Its key po<strong>in</strong>ts are the assumptions that language change is<br />

always an improvement with respect to a certa<strong>in</strong> parameter and that l<strong>in</strong>guistic structures can<br />

be situated on a graded scale of quality. In language change, improvements start with the<br />

structure that is worst accord<strong>in</strong>g to the preference law perta<strong>in</strong><strong>in</strong>g to the relevant parameter.<br />

But because parameters do not exist <strong>in</strong> isolation, improvements on one end can lead to a<br />

decrease <strong>in</strong> quality on another, which expla<strong>in</strong>s why change is not an improvement seen from<br />

every perspective. The strength of this concept is to elucidate and expla<strong>in</strong> language change as<br />

the implementation of preferences, which is especially valuable <strong>in</strong> account<strong>in</strong>g for<br />

bifurcational changes. Though a concrete sound change may not always be predicted,<br />

catalogues of possible changes make it possible that preference laws can be tested. As<br />

hypotheses the fundamental tenets of preference theory need not have a quantitative empirical<br />

basis, but are assumptions whose value lies <strong>in</strong> their application (see also Berg 1990: 570).<br />

<strong>Preference</strong>-like concepts have played a role <strong>in</strong> various l<strong>in</strong>guistic theories, most<br />

notably, Optimality Theory. But <strong>in</strong> contrast to most of these, they are not one-output models,<br />

which is precisely their strength. It seems that a good deal of the controversy preference laws<br />

have caused is due to an imperfect understand<strong>in</strong>g of their nature and their aims. We hope to<br />

have succeeded <strong>in</strong> remedy<strong>in</strong>g this to some degree.<br />

References<br />

Allen, Sidney (1953), Phonetics <strong>in</strong> Ancient India, London: Oxford University Press<br />

Allen, Sidney (1974), Vox Greaca: A guide to the pronunciation of Classical Greek, 2 nd<br />

edition, Cambridge: Cambridge University Press [first edition 1968]<br />

Bartsch, Renate and Theo Vennemann (1982), Grundzüge der Sprachtheorie, Tüb<strong>in</strong>gen:<br />

Niemeyer<br />

Bell, Alan (1977) The distributional syllable, <strong>in</strong> Juilland, Alphonse (ed.). L<strong>in</strong>guistic studies<br />

offered to Joseph Greenberg on the occasion of his sixtieth birthday.<br />

Saratoga/Kalifornien: Anma Libri, 249 -262<br />

Bell, Alan and Joan B. Hooper (1978), Issues and evidence <strong>in</strong> syllabic phonology, <strong>in</strong>: Alan<br />

Bell and Joan B. Hooper (eds), Syllables and Segments, Amsterdam: North Holland<br />

Publish<strong>in</strong>g Company<br />

Berg, Thomas (1990), Review of Vennemann (1988), Journal of L<strong>in</strong>guistics 26/2, 569-570<br />

Blev<strong>in</strong>s, Juliette (1995), The syllable <strong>in</strong> phonological theory, <strong>in</strong> John A. Goldsmith (ed.)<br />

(1995), The handbook of phonological theory, Cambridge, MA: Blackwell, 206-244<br />

Clements, G. N. (1990), The role of the sonority cycle <strong>in</strong> core syllabification, <strong>in</strong> Mary E<br />

Beckman & John K<strong>in</strong>gston (eds.) (1990), Papers <strong>in</strong> Laboratory Phonology I. Between<br />

the Grammar and Physics of Speech, Cambridge: Cambridge University Press, 283-333<br />

Coll<strong>in</strong>s, James. 1989. „Nasalization, lengthen<strong>in</strong>g, and phonological rhyme <strong>in</strong> Tolowa“.<br />

International Journal of American L<strong>in</strong>guistics 55, 326-340<br />

Dufter, Andreas (2003), Typen sprachrhythmischer Konturbildung, Tüb<strong>in</strong>gen: Niemeyer<br />

Dziubalska-Kołaczyk, Katarzyna (2001). Phonotactic constra<strong>in</strong>ts are preferences, <strong>in</strong>:<br />

Katarzyna Dziubalska-Kołaczyk (ed.), Constra<strong>in</strong>ts and <strong>Preference</strong>s, Berl<strong>in</strong>/New York:<br />

Mouton de Gruyter, 69-100.<br />

13


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

Evans, Nicholas (1995), Current Issues <strong>in</strong> Australian Languages, <strong>in</strong> John A. Goldsmith (ed.)<br />

(1995), The handbook of phonological theory, Cambridge, MA: Blackwell, 723-761<br />

Greenberg, Joseph H. (1978), Some generalizations concern<strong>in</strong>g <strong>in</strong>itial and f<strong>in</strong>al consonant<br />

clusters, <strong>in</strong> Joseph H. Greenberg (ed.), Universals of human language. Volume 2.<br />

Phonology. Stanford, CA: Stanford University Press, 243-279<br />

Hill, Eugen (2009), Die Präferenztheorie <strong>in</strong> der historischen Phonologie aus<br />

junggrammatischer Perspektive, Zeitschrift für Sprachwissenschaft 28, 231-263<br />

Holt, D. Eric (2003), Remarks on Optimality Theory and Language <strong>Change</strong>, <strong>in</strong>: D. Eric Holt<br />

(ed.), Optimality Theory and Language <strong>Change</strong>, Dordrecht: Kluwer, 1-30<br />

Hooper, Joan B. (1976), An <strong>in</strong>troduction to Natural Generative Phonology, New York:<br />

Academic Press<br />

Hurch, Bernhard and Richard A. Rhodes (eds.) (1996), Natural Phonology: the state of the<br />

art, Berl<strong>in</strong>/New York: Mouton de Gruyter<br />

Kl<strong>in</strong>genheben, August (1928), Die Silbenauslautgesetze des Hausa, Zeitschrift für<br />

E<strong>in</strong>geborenensprachen 18: 272-297<br />

Labov, William (1981), Resolv<strong>in</strong>g the Neogrammarian Controversy, Language 57: 267-308<br />

Labov, William (2010), Pr<strong>in</strong>ciples of L<strong>in</strong>guistic <strong>Change</strong>, Vol III. Malden, M.A.: Wiley-<br />

Blackwell<br />

Lutz, Angelika (1985), Die Worttrennung am Zeilenende <strong>in</strong> altenglischen Handschriften:<br />

Phonologische Betrachtungen zu Dieter Wetzels gleichnamigem Buch, Indogermanische<br />

Forschungen 90, 227-238<br />

Lutz, Angelika (1986), The syllabic basis of word division <strong>in</strong> Old English manuscripts,<br />

English Studies 6, 193-210<br />

Lutz, Angelika (1991), Phonotaktisch gesteuerte Konsonantenentwicklungen <strong>in</strong> der<br />

Geschichte des Englischen, Tüb<strong>in</strong>gen: Niemeyer<br />

<strong>Mailhammer</strong>, <strong>Robert</strong> (2009), Thoughts on the genesis and the development of syllable cut <strong>in</strong><br />

English, Anglia 127, 261-282<br />

Murray, <strong>Robert</strong> (1988), Phonological Strength and Early Germanic Syllable Structure,<br />

Munich: F<strong>in</strong>k<br />

Murray, <strong>Robert</strong> (1982), Consonant Cluster Development <strong>in</strong> Pāli, Folia L<strong>in</strong>guistica Historica<br />

111, 163-184<br />

Murray, <strong>Robert</strong> (2000), Syllable Cut Prosody <strong>in</strong> Early Middle English, Language 76, 617-654<br />

Murray, <strong>Robert</strong> and Theo Vennemann (1982), Syllable contact change <strong>in</strong> Germanic, Greek<br />

and Sidamo, Klagenfurter Beiträge zur Sprachwissenschaft 8, 321-349<br />

Murray, <strong>Robert</strong> and Theo Vennemann (1983), Sound change and syllable structure [:<br />

Problems] <strong>in</strong> Germanic phonology, Language 59, 514-528<br />

Pr<strong>in</strong>ce, Alan and Paul Smolensky (2002), Optimaliy Theory. Constra<strong>in</strong>t Interaction <strong>in</strong><br />

Generative Grammar, ROA version 8/2002 [http://roa.rutgers.edu/files/537-0802/537-<br />

0802-PRINCE-0-0.PDF]<br />

Restle, David and Theo Vennemann (2001), Silbenstruktur, <strong>in</strong>: Mart<strong>in</strong> Haspelmath, Ekkehard<br />

König, Wulf Oesterreicher and Wolfgang Raible (eds), Language Typology and<br />

Language Universals, II, Handbücher zur Sprach- und Kommunikationswissenschaft 20,<br />

Berl<strong>in</strong>/New York: Mouton de Gruyter, 1310-1336<br />

Selkirk, Elizabeth. 1982. Syllables, <strong>in</strong> Harry van der Hulst & Norval Smith (eds), The<br />

structure of phonological representations, Vol. 2, Dordrecht: Foris, 337-383<br />

Varma, Siddheswar (1929), Critical studies <strong>in</strong> the phonetic observations of Indian<br />

grammarians, London: Royal Asiatic Society [repr<strong>in</strong>ted 1961, New Delhi, India<br />

Munshiram Manoharlal Publishers Pvt. Ltd.]<br />

Vennemann, Theo (1972), On the theory of syllabic phonology, L<strong>in</strong>guistische Berichte 18, 1-<br />

18<br />

14


<strong>Mailhammer</strong>, Restle & Vennemann, <strong>Preference</strong> <strong>Laws</strong> <strong>in</strong> <strong>Change</strong><br />

Vennemann, Theo (1978), Universal syllabic phonology, L<strong>in</strong>guistische Berichte 18, 1-18<br />

Vennemann, Theo (1983a), Überlegungen zu e<strong>in</strong>er Theorie der l<strong>in</strong>guistischen<br />

Präferenzen, Klangenfurter Beiträge zur Sprachwissenschaft 9, 262-92.<br />

Vennemann, Theo (1983b), Causality <strong>in</strong> language change. Theories of l<strong>in</strong>guistic preferences<br />

as a basis for l<strong>in</strong>guistic explanations, Folia Liguistica Historica 6, 5-26.<br />

Vennemann, Theo (1986a). Neuere Entwicklungen <strong>in</strong> der Phonologie. Berl<strong>in</strong>: Mouton de<br />

Gruyter.<br />

Vennemann, Theo (1986b), Syllable based sound changes <strong>in</strong> Early Armenian, Annual of<br />

Armenian L<strong>in</strong>guistics 7, 27-43.<br />

Vennemann, Theo (1987), Muta cum Liquida: Worttrennung und Syllabierung im Gotischen,<br />

Zeitschrift für deutsches Altertum und deutsche Literatur 116, 165-204<br />

Vennemann, Theo (1988), <strong>Preference</strong> <strong>Laws</strong> for Syllable Structure and the Explanation of<br />

Sound <strong>Change</strong>, Berl<strong>in</strong>/New York: Mouton de Gruyter<br />

Vennemann, Theo (1989), Language change as language improvement, <strong>in</strong> V<strong>in</strong>cenzo Orioles<br />

(ed.), Modelli esplicativi della diacronica l<strong>in</strong>guistica: Atti del Convegno della Società<br />

Italiana de Glottologica, Pavia, 15-17 settembbre 1988, Pisa, 11-35 [repr<strong>in</strong>t <strong>in</strong> Charles<br />

Jones (ed.) Historical L<strong>in</strong>guistics: Problems and perspectives, 1993, London: Longman,<br />

310-344<br />

Vennemann, Theo (2000a), Triple-cluster reduction <strong>in</strong> Germanic: Etymology without sound<br />

laws? Historische Sprachforschung 113, 239-58.<br />

Vennemann, Theo (2000b), From quantity to syllable cuts: On so-called lengthen<strong>in</strong>g <strong>in</strong> the<br />

Germanic languages, Italian Journal of L<strong>in</strong>guistics 12, 251-282<br />

15

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!