A Beginner's guide to Slavic accentuation

[Zekoslav]

Slavic accentuation (the free, mobile accentuation seen in South and East Slavic languages) is extremely complex and extremely interesting, and it seems to be a good time to share my knowledge about it. Should I develop it properly, this thread will be a good resource for anyone interested in Slavic languages, natural or constructed.

I am a native speaker of a Slavic language (Kajkavian dialect of Croatian) and a soon-to-be professional linguist - I took a year's course on diachronic Slavic accentuation, which included reading some sizeable literature not easily acquired on the Internet (usually because it's not in English). I think I got a decent understanding of the topic, so that I can make a good summary of the literature I've read.


An Introductory warning:

The development of Slavic accentuation is not well understood - it's impossible to explain it solely by sound laws, due to it's intimate relation with morphology. The further one goes, things become less and less clear - when it comes to the development of (Balto-)Slavic accentuation from the Proto-Indo-European one, there's multiple theories per scientist...

Because of this, I won't be able to post a complete list of sound changes from PIE. to modern Slavic languages - instead, I will focus on explaining how the system works from a synchronic perspective, with some diachrony to tie the different languages together. Without further ado...


The basics of Slavic accentuation:

Intro:

What are the basics of Slavic accentuation and how does it differ from more familiar systems of free, mobile accent?

Other Indo-European languages also possess a free, mobile accent, whether they inherited it from PIE. (Greek, Vedic Sanskrit) or developed it independently (the Romance languages, English). Due to their familiarity, the Romance languages will make a good comparison.


Definitions:

I speak of accent, because both stress and pitch accent behave similarly in the aforementioned languages.

• A free accent is one whose placement is not predictable from the phonological properties of the word. This means that two words whose phonological shape is the same can have a different placement of accent - in other words, placement of accent is phonemic (for this purpose, it doesn't matter whether the difference is lexical or grammatical).

eg. Zagreb Croatian /ˈplakat/ "to cry" vs. /plaˈkat/ "poster", /ˈnovine/ "newspaper" vs. /noˈvine/ "novelties", Spanish /ˈamo/ "I love", /aˈmo/ "he/she/it loved"

• A mobile accent is one whose placement differs between different morphological forms of the same word. This can happen in both derivational and inflectional morphology - although English lacks mobility of the latter type (Therefore, mobile accent withing an inflectional paradigm is somewhat exotic to English speakers).

eg. English <ˈmobile> vs. <moˈbility>, Zagreb Croatian /ˈime/ "name" vs. /iˈmena/ "names", Spanish /ˈamo/ "I love", /aˈmamos/ "we love"


A comparison of Romance and Slavic accent systems:

So, English, Romance and Slavic all have a free, mobile accent system. What are the differences?


1. Scope of accentual mobility.

Due to historical reasons, in Romance, accentual mobility is limited - Latin accent was limited to the last three syllables of the word, which was further limited in Romance by syncopes and apocopes. Therefore, when accent moves within the paradigm, it will move one or two syllables:

eg. Spanish /espeˈθifiko/ "specific" vs. /espeθiˈfiko/ "I specify" vs. /espeθifiˈko/ "he/she/it specified


Again due to historical reasons, no such restriction exists in Slavic, and when the accent moves withing the paradigm, it can skip multiple syllables:

eg. archaic Čakavian /ˈnebo/ "sky" vs. /nebeˈsǎː/ "heavens", /ˈnagovoriːl/ "he persuaded" vs. /nagovoriːˈla/ "she persuaded"


Another difference is in the behavior of proclitics: in Romance, proclitics are never accented, while in Slavic, they can be accented (enclitics are never accented in modern Slavic). Furthermore, when the accent moves withing a paradigm, it can skip multiple proclitics:

eg. archaic Čakavian /ˈna ruːku/ "onto the hand" vs. /na ruːˈki/ "on the hand", /ˈi na ruːku/ "and onto the hand" vs. /i na ruːˈki/ "and on the hand"


2. Predictability of accentual mobility.

While it is not predictable from the standpoint of phonology, in the Romance languages, placement of accent, including it's mobility, can be predicted from the standpoint of grammar and lexicon. The rules for accenting Spanish (regular) verbs are quite simple, with each grammatical form having it's characteristic accentual properties, regardless of the conjugation class of the verb - some suffixes are unaccented, and others are accented. The same is true for derivational suffixes, some being accented, and others unaccented.


/ˈamo/ "I love" vs. /aˈmamos/ "we love" vs. /aˈmaba/ "I loved" vs. /aˈmabamos/ "we loved" vs. /aˈme/ "I('ve) loved" vs. /amaˈ re/ "I'll love"

/ˈtemo/ "I fear" vs. /teˈmemos/ "we fear" vs. /teˈmia/ "I feared" vs. /teˈmiamos/ "we feared" vs. /teˈmi/ "I('ve) feared" vs. /temeˈre/ "I'll fear"

/ˈparto/ "I split" vs. /parˈtimos/ "we split" vs. /parˈtia/ "I was splitting" vs. /parˈtiamos/ "we were splitting" vs. /parˈti/ "i('ve) split" vs. /partiˈre/ "I'll split"


This will be called a category-based mobile accent, because the grammatical category of a word (inflectional or derivational) usually predicts it's accent pattern - as can be seen above, 1. sg. present is accented on the stem, while 1. pl. present is accented on the person-number suffix, the imperfect tense is always accented on the tense suffix, while the preterite and future tenses are always accented on the person-number suffix, regardless of the conjugation class of the verb and the exact form of the grammatical morpheme.

On the other hand, in Slavic, the grammatical category of a word doesn't predict it's accent pattern, as can be seen in these three verbs in archaic Kajkavian, all belonging to the same conjugational class, but having different accent patterns (I've chosen archaic Kajkavian because it shows that not only the placement of accent, but also length and tone, can change, seemingly at random).


/ˈpaziti/ "to watch" vs. /ˈpâːzim/ "I watch/I'm watching" vs. /pâːzimo/ "we watch/we're watching" vs. /ˈpazil/ "he watched" vs. /ˈpâːzila/ "she watched" vs. /ˈpazi/ "watch out!"

/braˈniti/ "to protect" vs. /ˈbrǎːnim/ "I protect/I'm protecting" vs. /ˈbrǎːnimo/ "we protect/we're protecting" vs. /braːˈnil/ "he protected" vs. /ˈbrǎːnila/ "she protected" vs. /braːˈni/ "protect!"

/saˈditi/ "to plant" vs. /saˈdǐːm/ "I plant/I'm planting" vs. /sadiːˈmo/ "we plant/we're planting" vs. /ˈsadil/ "he planted" vs. /saˈdǐːla/ "she planted" vs. /saːˈdi/ "plant!"


Also, different morphemes expressing the same grammatical category can have different accentual properties, as can be seen in this list of Russian stress patterns of nouns - different cases have different accentual properties depending on the declension, and all patterns are equally regular. Something similar does exist in Spanish, where a group of preterites (so-called "strong preterites") has a distinct accent pattern from the one shown above, but these are considered irregular.

Such an accent system, where the accentual behavior of both roots (roots of same inflectional class can have different accent pattern) and suffixes (suffixes expressing the same grammatical category can have different accentual properties) is free from the grammatical category they belong to is called a paradigm-based mobile accent, because each word belongs to one of unpredictable (lexically specified) accent patterns, or paradigms.

This is the most salient feature of the Slavic accent system, and the one that baffles people the most (both professionals and non-professionals). In fact, Slavic languages are the only Indo-European languages that posses a mostly paradigm-based accent system - the accent system of Romance is mostly category-based (with the exception of irregular preterites), while the accent system of PIE, Greek and Vedic Sanskrit is mixed - it's, to simplify, paradigm-based in nouns, and category-based in verbs.

In the next post, I will show how to analyze the working of a paradigm-based accent system - such an analysis works on a typological level and can be applied to many unrelated languages, not just Slavic.

Feel free to comment and ask if anything isn't clear. I hope this grows into a fruitful discussion .

[Omzinesý]

Very well written introduction. You seem to be very thorough with terms and such which is very good.
Cannot wait for seeing the real description.

[Zekoslav]

Very well written introduction. You seem to be very thorough with terms and such which is very good.
Cannot wait for seeing the real description.

Thanks! Proper definition of terminology is my own contribution, since usually different linguistic schools have different terminologies and reading different papers can confuse uninitiated readers. Since this is a complex topic, I think it would be good for everyone to know what exactly I'm talking about .

I've got one question - I "assumed" everyone is more or less familiar with Spanish when writing the comparison of Romance and Slavic accent systems (to say nothing of being already rather tired ), so I didn't really write out any Spanish verb conjugations for comparison. Do you think I should edit my post and do it (to show that in Spanish all regular verbs have the same accent pattern)?

Edit: I've revised the post anyway, but would still like to know if it's improved.

[Zekoslav]

A phonological analysis of free, mobile accent systems.

The way a free, mobile accent system (of the Slavic type) works is in fact quite simple -it's a weight-based accent system, such as that of Latin, Sanskrit or Classical Arabic, but instead of being based on the (surface) weight of syllables, it is based on the (underlying) weight of morphemes.

However, like systems of phonological weight, systems of morphological weight tend to become more complicated with time due to sound changes (which then tends to lead to simplifications - the development of the Romance accent system is a good example of this, which I will detail in the chapter on diachrony). Because of this, I will demonstrate how it works on the example of (Pre-)Proto-Slavic, when it was still rather simple (this might be an artifact of the comparative method - there might be hidden complexities, which we will ignore for the purposes of this introduction).

• A quick look at (Pre-)Proto-Slavic

Pre-Proto-Slavic was quite different from modern Slavic languages - fortunately, that means it had a very simple phonology. It was also quite archaic - for a comparison, it was the Classical Latin to modern Slavic's French (Lithuanian would be modern Slavic's Italian).

It had a consonant inventory of /p t k b d g s z x m n l r w j/ and a vowel inventory of /i iː e eː a aː u uː/, with numerous diphthongs: in Pre-Proto-Slavic, any combination of a vowel (long or short) and a sonorant (not only /w/ and /j/, but also /m/, /n/, /l/ and /r/) counted as a dipthong, because for the purpose of accent, it patterned together with long vowels.

In Pre-Proto-Slavic, long vowels and dipthongs had a phonation contrast traditionally called acuteness (that is, there was a contrast between acute and non-acute long vowels and dipthongs). What it's phonetic realization is is unclear - in Latvian, it's preserved as glottalization (the so-called broken tone), while it's reflexes in Lithuanian and Slavic could develop from both glottalization and creaky voice.

Acuteness was originally completely separate from accent, but interacted with it in complex ways (only one such interaction, Hirt's Law, belongs to the Proto-Balto-Slavic stage), which is the main reason why I choose Pre-Proto-Slavic instead of a modern Slavic language -it's simpler and more regular. Don't worry, though - I will talk about modern Slavic languages in a future post.

Because it's phonetic realization is unclear (and because IPA's combining diacritics don't combine well in most fonts), acuteness will be represented with a circumflex - the diacritic used to represent Latvian's broken tone.

• An analysis of the Pre-Proto-Slavic accent system

1. More terminology

A system of accent based on morphological weight divides the morphemes of a language into two classes, called accented and unaccented by generativists and dominant and recessive by slavists (especially the Moscow School). These two correspond to heavy and light syllables in systems of accent based on syllable weight.

These terms are unfortunate (I list them because when reading published research, you will most commonly see these terms), because underlying accent doesn't always correspond to surface accent, so an "accented" morpheme may end up unaccented on the surface, and an "unaccented" morpheme may end up as accented - to say nothing of the fact that different schools of linguistics have different terminology (as above).

So, I will use the terms strong and weak to refer to these two classes of morphemes, restricting the terms accented and unaccented to refer exclusively to the surface realization of accent as stress or pitch (depending on the language).


2. The good stuff

When deriving the accent of a word, it's necessary to a) know the accentual class of the morphemes it consists of and b) know the rules which assign the accent to one of these morphemes. For simplicity and clarity, I will explain how this works in Pre-Proto-Slavic on bimorphemic words consisting of monosyllabic morphemes. Polymorphemic words, polysyllabic morphemes and clitics will be explained in the next post.


Rules of accent assignment in Pre-Proto-Slavic:

1. If the word contains one or more strong morphemes, the leftmost strong morpheme receives an accent. This accent is realized as a high tone, which will be marked as <ˈ>.

2. If the word contains no strong morphemes, the initial (=leftmost) syllable receives an accent. This accent is realized as a low tone, which will be marked as <ˌ>.

• Note: Words containing no strong morphemes are considered to be underlyingly unaccented, and the initial low tone is the realization of the underlying lack of accent (this is similar to how in Japanese, underlyingly unaccented words lack the drop in pitch characteristic of underlyingly accented words). There is some evidence that they were originally truly unaccented (clitics), receiving the initial low tone only in isolation! Because of this, in slavistics, underlyingly unaccented words are called enclinomena.

These rules apply equally to roots, derivational affixes and inflectional affixes - this equal treatment of all morphemes is characteristic of the (Pre-)Proto-Slavic accent system, and only partially preserved in modern Slavic languages. It's also the defining characteristic of a paradigm-based accent system - if affixes are always stronger than roots (which is common, cross-linguistically, for derivational affixes), the result is a category-based accent system. I will talk about this in a future post.

Let's take two nouns that decline according to the same declension pattern - their roots are /gen-/ "woman" which is strong (+), and /pent-/ "heel" which is weak (-). Let's decline them in the nominative and the accusative case - their suffixes are /-âː/ "Nom.Sg.", which is strong (+), and /-aːn/ "Acc.Sg.", which is weak (-).

Combining these morphemes, in the nominative case we get /gen(+)-âː(+)/ and /pent(-)-âː(+)/, and in the accusative case we get /gen(+)-aːn(-)/ and /pent(-)-aːn(-)/. How are these two words accented?


1. Applying the first rule, we see that /ˈgenâː/ "woman-Nom.Sg." and /ˈgenaːn/ "woman-Acc.Sg." are accented on the first syllable, while /penˈtâː/ "heel-Nom.Sg." is accented on the second syllable. The accent is a high tone in all cases.

2. Applying the second rule, we see that /ˌpentaːn/ "heel-Acc.Sg" is accented on the first syllable, and that the accent is a low tone.


By applying the rules to different combinations of strong and weak morphemes, a pattern has emerged, which can be seen when we compare the two words' morphological paradigms:

Code: Select all

Nom.  /ˈgenâː/    / penˈtâː/
Acc.  /ˈgenaːn/   /ˌpentaːn/

In the first word's morphological paradigm, the accent stays on the same syllable in all of it's word-forms. Words with this accent pattern belong to the immobile accent paradigm.

In the second word's morphological paradigm, the accent moves between the initial syllable and the affix depending on the exact word-form. Words with this accent pattern belong to the mobile accent paradigm.


An accent paradigm, not to be confused with a morphological paradigm, is therefore a class of words whose accent follows the same general pattern. The existence of accent paradigms is why the accent system of the Slavic type is called a paradigm-based accent system. More precisely, in a paradigm-based accent system, a word's accent paradigm must not depend on it's grammatical category, unlike in a category-based accent system where it's the opposite (keeping in mind that most natlangs have mixed systems). As I've already hinted, the difference lies in the relative strength of roots and affixes, which I will talk about in a future post.

Since accent paradigms are independent of grammatical categories, words of all classes (nouns, adjectives, verbs...) can belong either to the immobile or the mobile accent paradigm. However, grammatical category does play a role in dividing these accent paradigms into smaller sub-paradigms - since different morphemes expressing the same grammatical category can have different accentual properties (for example, while the Gen. Sg. of feminine a-stems is strong, the Gen. Sg. of feminine i-stems is weak.), different declension or conjugation classes have different patterns of accent movement in the mobile accent paradigm - that is, they have different accentual curves.

These are the fundamentals of the Slavic accent system. The system of accent paradigms was quite healthy in Pre-Proto-Slavic times, probably because the opposition of high and low pitch accents easily differentiated mobile and immobile accent paradigms, but that changed drastically in modern Slavic languages.

• Until next time...

In the next post, I will be exploring Pre-Proto-Slavic accent in more detail, but if you're feeling impatient and courageous, here's two papers which you can read in the meantime (which will probably make the next post useless):

Compositional vs. Paradigmatic Approaches to Accent and Ablaut by Paul Kiparsky, which analyses the Sanskrit and PIE. accent systems using the same approach as laid out here.

• Note: There might be some terminological confusion when reading the paper - it's important to differentiate between morphological and accent paradigms. The (morphological) paradigmatic approach criticized in the paper is criticized for conflating the two, which is good enough for analyzing a category-based accent system, but not for analyzing a (accentual) paradigm-based accent system.

Structure and Stress in the Phonology of Russian, by Janis. L. Melvold, which analyses the Russian accent system using the same approach.

Edit: Edited for clarity.

[Clio]

Thank you for posting this! I don't speak any Slavic language and don't know much about any, but everything is comprehensible and interesting.

I have one question, since I want to be sure I don't read too much into your examples. The nominative singular ending happens to be both strong and acute, while the accusative singular is neither. Is it just coincidence that the two qualities line up in these examples, and were there Balto-Slavic case endings which were week and acute or strong and non-acute? I assume so but want to be sure.

[Ælfwine]

I’m trying to wrap my head around this.

A few questions:

• Is there a way to predict what word has what accent, if nothing is known about the accent?
• Are initial “strong” morphemes always have immovable accents?

Otherwise, it looks great Zeko!

[gestaltist]

Love this thread. Very informative. :)

[Zekoslav]

Thank's everyone for your feedback!

@Clio: Yes, it's just a coincidence (I choose nominative and accusative singular so I don't overwhelm the reader with whole morphological paradigms). For example, the a-stem genitive singular /-aːs/ is strong and non-acute, while the a-stem accusative plural /-âːns/ is weak and acute.

As was written, acuteness was originally completely separate from accent - the exact rules by which it has come to be are unclear, but acute syllables usually originate from vowels followed by a laryngeal in PIE., and non-acute vowels usually originate from vowel contraction (what's unclear is what happens to the original long vowels - they sometimes become acute, and sometimes non-acute, probably depending on the following consonant).

(Incidentally, these rules suggest that the laryngeals were first lost intervocalically and later in coda and onset positions, similar to Proto-Germanic *h in English.)

I recommend you to read Kiparsky's papers (especially the one I've posted), since they will help you develop Getic accent.


@Ælfwine: Concerning the first question, no - in a system of free accent, accent is always unpredictable. Each morpheme is lexically specified for accent and it has to be remembered by both the learners and the researchers. Fortunately, a good dictionary will always mark accent (even Wiktionary does it, which is very helpful). Concerning the second question, yes - words whose initial morphemes (prefixes or roots) are strong always belong to the immobile accent paradigm, at least initially - once the Proto-Slavic opposition of high and low pitch accents (which helps differentiate immobile and mobile accent paradigms) breaks down, so-called "secondary mobility" can arise - words switch from the immobile to the mobile accent paradigm.

But, for the Proto-Slavic stage a strict rule can be established, that a mobile word can never be derived from an immobile word, while the opposite can and does happen. This follows from the rules of accent I posted and I intend to talk about it in a future post.


@Gestaltist: Thank you! Keep following the thread, it will get even more interesting.

[Zekoslav]

Accent of longer words pt. 1

The rules of accent assignment detailed in the last post apply to words containing any number of morphemes. In this post, I will demonstrate how this works with words containing three morphemes (a root, a derivational suffix and an inflectional suffix).


1. Classes of strong morphemes

Most derivational affixes in Pre-Proto-Slavic work exactly as roots and inflectional affixes work - they are either strong or weak, and usual rules apply to them. Some strong derivational affixes, however, have special properties - instead of being accented according to rule no. 1, they are pre-accenting or post-accenting.

As their names suggest, pre-accenting strong affixes produce an accent on the preceding syllable, and post-accenting strong affixes produce an accent on the following syllable, in contrast to plain strong suffixes, which produce an accent on their own syllable (I will call them self-accenting affixes). Pre-accenting strong morphemes will be designated as (<+), and post-accenting strong morphemes will be designated as (+>).

Modern Slavic languages possess all three types of affixes - both derivational and inflectional affixes can be self-accenting, pre-accenting or post-accenting (there also exists a number of post-accenting roots, which I will talk about in a future post concerning diachrony).

Let's take archaic Kajkavian (the Slavic language I'm most familiar with with an accent system archaic enough to demonstrate these types of suffixes) for example, and consider these words:

• post-accenting suffixes:

/ˈbôːg/ "God-Nom.Sg.", /ˈboga/ "God-Gen.Sg."

/boˈʒiʧ/ "Christmas-Nom.Sg.", /boˈʒiʧa/ "Christmas-Gen.Sg."

/boˈgeʦ/ "poor-Nom.Sg.", /bokˈʦa/ "poor-Gen.Sg."


They are all derived from the same root /bog-/ (-) meaning "god" or "richness" (cognate with Vedic Bhaga, the god of richness), and inflected in the nominative and the genitive cases, both of whose suffixes are weak (-) in the relevant declension pattern.


The first two contains self-accenting strong morphemes and weak morphemes:

- The word "God", containing no additional morphemes, therefore has an initial accent according to rule no. 2 (in words containing no strong morphemes, the initial syllable receives an accent).

/bog(-)-Ø(-)/, /bog(-)-a(-)/


- the word "Christmas", containing the morpheme /-iʧ-/ (+), therefore has an accent on that syllable according to rule no. 1 (in words containing strong morphemes, the leftmost strong morpheme receives an accent)

/boʒ(-)-iʧ(+)-Ø(-)/, /boʒ(-)-iʧ(+)-a(-)/


The third one contains post-accenting strong morphemes and weak morphemes:

- the word "poor", containing the morpheme /-Eʦ-/ (+>), is accented on that syllable in the nominative case, and on the following syllable in the genitive case. This is because a post-accenting morpheme only produces an accent on the following syllable if there is a syllable for it to produce an accent on, otherwise it behaves as a self-accenting morpheme.

/bog(-)-eʦ(+>)-Ø(-)/, /bok(-)-ʦ(+>)-a(-)/

• pre-accenting suffixes:

/ˈimɛ/ "name-Nom.Sg.", /ˈimɛna/ "name-Gen.Sg.", /iˈmɛ̌ːna/ "name-Nom/Acc.Pl."


This word's bisyllabic stem is weak (-), as are the suffixes for the nominative singular and the genitive singular cases in the relevant declension:

/imɛ(-)-Ø(-)/, /imɛn(-)-a(-)/


The suffix for the nominative and accusative plural cases is, on the contrary, pre-accenting (<+), and produces an accent on the preceding syllable:

/imɛn(-)-a(<+)/


Pre-Proto-Slavic, however, had no post-accenting suffixes, and pre-accenting suffixes were limited to being derivational. This can be explained by the fact that post-accenting suffixes break rule no. 1 - that the first strong morpheme in a word receives the accent. Pre-accenting suffixes can be analyzed as following rule no. 1 if one analyzes them as changing the accentual class of the preceding morpheme from weak to strong (which also explains why they were limited to being derivational, because changing accentual properties of morphemes is limited to derivation).

As an example, consider this Pre-Proto-Slavic word, formed from the morphemes /dalb-/(-) "to hollow", /-t-/(<+) "nominalizer suffix", and /-an/(-) "Nom/Acc.sg.Neut.":


/dalb(-)-t(<+)-an(-)/ > /dalb(+)-t(<+)-an(-)/ > /ˈdalptan/ "chisel-Nom/Acc.Sg."


2. Accent assignment in polymorphemic words

With the exception of pre-accenting derivational suffixes, accent assignment in polymorphemic words follows the same two rules as does accent assignment in bimorphemic words (that is to say, it is unpredictable from the rules of derivation and inflection themselves, and the strength of all individual morphemes has to be taken into account). To demonstrate this I will use the following Pre-Proto-Slavic morphemes:


- roots: /gen-/(+) "woman", /gar/(-) "mountain"

- derivational suffixes: /-îːk-/(+) "deminutive.Fem.", /-isk-/(-) "relational adjective"

- inflectional affixes: /âː/(+) "Nom.Sg.Fem.", /aːn/(-) "Acc.Sg.Fem."


Bimorphemic combinations:


1. /gen(+)-âː(+)/, /gen(+)-aːn(-)/ > /ˈgenâː/ "woman-Nom.Sg.", /ˈgenaːn/ "woman-Acc.Sg."

2. /gar(-)-âː(+)/, /gar(-)-aːn(-)/ > /gaˈrâː/ "mountain-Nom.Sg.", /ˌgaraːn/ "mountain-Acc.Sg."


Polymorphemic combinations:

3. /gen(+)-îːk(+)-âː(+)/ /gen(+)-îːk(+)-aːn(-)/ > /ˈgenîːkâː/ "woman-dem.-Nom.Sg.", /ˈgenîːkaːn/ "woman-dem.-Acc.Sg."

4. /gen(+)-isk(-)-âː(+)/ /gen(+)-isk(-)-aːn(-)/ > /ˈgeniskâː/ "woman-poss.-Nom.Sg.", /ˈgeniskaːn/ "woman-poss.-Acc.Sg."


5. /gar(-)-îːk(+)-âː(+)/ /gar(-)-îːk(+)-aːn(-)/ > /gaˈrîːkâː/ "mountain-dem.-Nom.Sg.", /gaˈrîːkaːn/ "mountain-dem.-Acc.Sg."

6. /gar(-)-isk(-)-âː(+)/ /gar(-)-isk(-)-aːn(-)/ > /garisˈkâː/ "mountain-poss.-Nom.Sg.", /ˌgariskaːn/ "mountain-poss.-Acc.Sg."


The result of the application of accent assignment rules is that some words (2, 6) are mobile, and some are immobile (1, 3, 4, 5). When observing the accentual relationship between underived and derived words, a rule can be seen: immobile words can be derived from mobile words, but mobile words can't be derived from immobile words.

- immobile to immobile: 1 > 3, 4

- mobile to immobile: 2 > 5

- mobile to mobile: 2 > 6

This follows from the rule no. 1, that the first strong morpheme in the word receives the accent - strong roots remain strong whether they are suffixed with strong or weak suffixes (immobile to immobile), while weak roots remain weak whether they are suffixed with strong (mobile to immobile) or weak (mobile to mobile) suffixes.

Therefore, to find which accent paradigm a word belongs to, one has to check the strength of it's morphemes from left to right: first roots, then derivational affixes if the root is weak, then inflectional affixes if derivational affixes are weak - a word will belong to the mobile accent paradigm only if it's stem contains no strong morphemes.

This can be hard to wrap one's head around (that every morpheme has to be checked individually, starting from the beginning of the word). Modern Slavic languages preserve only traces of this system, and most derivational affixes impose their accent class (strong or weak) onto the derived stem, ignoring the accent class of the root. Affixes which do this are called dominant, and affixes which don't do that are called recessive (except by the Moscow School, where these term refer to strong and weak affixes).

Proto-(Balto-)Slavic, on the contrary, preferred to classify accent paradigms of derived words by root rather than by derivational suffix - in fact, all derivational suffixes that were dominant in PIE. became recessive in Proto-(Balto-)Slavic!


3. Conclusion

To summarize what we've learned that makes the (Pre-)Proto-Slavic accent system so interesting when compared to other morphological weight-based accent systems:

1. It has only two classes of morphemes, lacking post-accenting affixes entirely and limiting pre-accenting affixes to derivation.

2. It lacks dominant affixes, which completely separates accentual categories from grammatical categories - this is, therefore, the essential feature of a paradigm-based accent system.

3. Two rules of accent assignment are enough to assign the accent to any combination of morphemes.

4. It has underlyingly unaccented words.


I originally wanted to treat clitics in this post as well, but it turned out to be too long. Clitics, as well as the loose ends of the Slavic accent system, will be treated in the next post.

[Aevas]

I haven't taken the time to read through this thread yet, but I wanted to say thank you for making it. I look forward to learning something when I finally get around to it!

[Zekoslav]

I haven't taken the time to read through this thread yet, but I wanted to say thank you for making it. I look forward to learning something when I finally get around to it!

Thank you! I've been rather busy recently, but I intend to continue the thread. In the last post I tried to explain too many things at once, so I think I didn't do them justice. I'll try to present each of the topics properly in the next few posts.

[Zekoslav]

More about accent paradigms in derivation

I'm back! I took me a lot of time to think through how to continue this thread, and the beginning of college didn't help with finding time to do it. But, now the Beginner's guide to Slavic accentuation is officially back in action!


1. Headedness of weight-sensitive systems

The explanation of the behavior of accent paradigms in derivation is heavily based on this chapter in WALS. I highly recommend you to read it before reading this post.

Weight-sensitive accent systems, whether based on syllable weight like Latin or on diacritic (i.e. underlying) weight like Slavic, can be either left-headed or right-headed. This parameter defines which syllable in a sequence of syllables of the same weight will be accented (syllables of higher weight, i.e. heavy or strong, are always accented in preference to syllables of lesser weight, i.e. light or weak). In a left-headed system, it will be the leftmost such syllable, while in a right-headed system it will be the rightmost one.

A language can have one parameter set for heavy/strong syllables, and another for light/weak syllables. Slavic is left-headed for both strong and weak syllables, Latin and Sanskrit are right-headed for heavy syllables and left-headed for light syllables, while Sogdian is left-headed for heavy syllables and right-headed for light syllables.

In practice, this means that in order to find the accented syllable, in a left-headed system such as Slavic one has to check syllables for weight starting from the left edge of the word, while in a right-headed system such as Latin one has to check them starting from the right edge of the word.

Most of the differences between the Slavic and Romance accent systems I listed in the first post actually arise because one is left-headed and the other is right-headed for it's heavy syllables (the rest arise because the Slavic system is unbounded, while the Romance one is bounded) - this will be a topic for a future post.


2. How does morphology fit into all of this?

a. Inflection

Slavic languages, like most Indo-European languages, are inflectional/fusional and suffixing. This means that within a paradigm, only the last morpheme will differentiate inflected word-forms from one another, and therefore only the last morpheme will differentiate the accent of these inflected word-forms.

Since strong morphemes are always accented in preference to weak morphemes, and since Slavic accent system is left-headed for both strong and weak morphemes, this means that:

• a) words whose stem contains at least one strong morpheme will belong to the immobile accent paradigm

• b) words whose stem contains no strong morphemes will belong to the mobile accent paradigm

This gives us the condition for accentual mobility: In order for a word to belong to the mobile accent paradigm, it's stem must contain weak morphemes only.

b. Derivation

In derivation, the same process is repeated again with every additional derivational suffix: if the root is strong, all derived words will be immobile, if the root is weak, derived words will be immobile if derived with a strong suffix and mobile if derived with a weak suffix; if the resulting stem is strong, all further derived words will be immobile, if the resulting stem is weak, further derived words will be immobile if derived with a strong suffix and mobile if derived with a weak suffix; and so on.

An interesting consequence of this is that, statistically, derived words are more likely to be immobile, while underived words are more likely to be mobile.

e.g. Proto-Slavic:

(Note: I decided to represent acute vowels with the little superscript glottal stop, to represent their hypothesized phonetic realization)


sủ:n- (-) "son" + -aw- (-) "adj.poss."> sủ:n-aw- (mob.) "son's" : ˌsủ:nawas "son's-Nom.Sg.masc.", sủ:naˈwả: "son's-Nom.Sg.fem."

sủ:n-aw- (mob.) "son's" + -isk- (-) "adj.rel." > sủ:n-aw-isk- (mob.) "sonly" : ˌsủ:nawiskas "sonly-Nom.Sg.masc.", sủ:nawisˈkả: "sonly-Nom.Sg.fem."

sủ:n-aw- (mob.) "son's" + -ik- (+) "dimin." > sủ:n-aw-ik- (immob.) "a type of cousin" : sủ:naˈwikas "...-Nom.Sg.", sủ:naˈwikả: "...-Inst.Sg."

sủ:n-aw-ik- (immob.) "..." + -isk- (-) "adj.rel." > sủ:n-aw-ik-isk- (immob.) "...-ly" : sủ:naˈwikiskas "...-ly-Nom.Sg.masc.", sủ:naˈwikiskả: "...-ly-Nom.Sg.fem."


The observation that from mobile words both mobile and immobile words can be derived, but from immobile words only immobile words can be derived is simply a practical summary of the principles listed above.

Next post: clitics!

[kanejam]

Just bumping this thread to say that it's great! Very informative, interesting, wonderfully laid out and easy to follow. I would love to see some more, if Zekoslav is up to it (it looks like they are still active on this site ).

[Zekoslav]

Just bumping this thread to say that it's great! Very informative, interesting, wonderfully laid out and easy to follow. I would love to see some more, if Zekoslav is up to it (it looks like they are still active on this site ).

Thanks! I'm glad to hear it's easy to follow: this is a complicated topic and I had quite a headache while writing this (most of this due to my own, ahem, perfectionism)... I'd have to re-read this thoroughly in order to continue it in an orderly manner.

I mostly gave up because of a lack of feedback and the resulting difficulty in deciding what to do next. So far it looks like people are interested in the diachronic side of things... this is quite a can of worms and can't be properly understood without understanding the synchronic side, thankfully I've already mostly finished this, and there's only some loose ends left to take care of.

[Clio]

If it helps, I was really looking forward to a post on clitics, but I didn't say anything since I didn't want to seem pushy. I'm going to read through this thread again as well, hoping to be prepared for wherever it leads next.

[Zekoslav]

If it helps, I was really looking forward to a post on clitics, but I didn't say anything since I didn't want to seem pushy. I'm going to read through this thread again as well, hoping to be prepared for wherever it leads next.

Oh, don't worry, clitics are still coming next! Personally, I had a burnout writing this thread and then decided to concentrate on other stuff (including my two master's theses...), so this time updates will probably come less frequently (I'll try to avoid another hiatus like this, however.).

[Zekoslav]

Clitics

It's time to finally deal with the long awaited clitics!


1. General stuff

A clitic is, as Wikipedia nicely puts it, a syntactically independent but phonologically dependent morpheme. For the purposes of accent, clitics behave the same as bound morphemes, and adding a clitic to a word is the same as adding an affix to the word’s stem. Since bound morphemes can be strong or weak, it’s easy to deduce that clitics can also be strong or weak… however, this deduction is incomplete in that clitics being strong or weak is not a lexical property like for bound morphemes, but a morphosyntactic property: proclitics all behave as weak morphemes and enclitics all behave as strong morphemes.


2. Proclitics

Recall the two rules of accent assignment: 1. if there are any strong morphemes in the word, the first one is accented with a high tone, and if there are no strong morphemes in the word, the initial syllable is accented with a low tone. Since proclitics are weak morphemes and since they come before the word they depend on, this means that if the word they depend on contains any strong morphemes, the proclitic will be unaccented, but if the word contains no strong morphemes, the proclitic will be accented. In other words, adding a proclitic to a word with an underlyingly accented high tone does nothing, but adding a proclitic to a word with an underlyingly unaccented low tone will shift the accent from the initial syllable of the word to the initial syllable of the proclitic, skipping any intermediary syllables:

Code: Select all

ˈkảrw-a:n ‘cow-A.sg.’  >  pra ˈkảrw-a:n ‘for cow-A.sg.’,   perdan ˈkảrw-a:n ‘in.front cow-A.sg.’,  perdan ˈkảrw-ajả:n ‘in.front cow-I.sg.’
  +   –                     –   +   –                       –       +   –                           -       +     +

ˌgảlw-a:n ‘head-A.sg.’  >  ˌpra gảlw-a:n ‘for head-A.sg.’,  ˌperdan gảlw-a:n ‘in.front head-A.sg.’,  perdan gảlw-aˈjả:n ‘in.front head-I.sg.’
  –   –                       –  –   –                        –      –   –                            –      –      +

The accent patterns this leads to are the following: with words belonging to the immobile accent paradigm (strong stem), proclitics are never accented, with words belonging to the mobile accent paradigm (weak stem), proclitics are unaccented if the suffix is strong, and accented if the suffix is weak. Often, as shown in the examples the same preposition is unaccented when used with one case, and accented when used with another, since some case endings are strong and others are weak.


3. Enclitics

Since enclitics are strong morphemes, they will be accented with a high tone if they also happen to be the first strong morpheme in the word. However, since they come after the word they depend on, this will only happen if there are no strong morphemes in the word they depend on. In other words, same as with proclitics, adding an enclitic to a word with an underlyingly accented high tone does nothing, but adding an enclitic to a word with an underlyingly unaccented low tone will shift the accent from the initial syllable of the word to the final syllable of the enclitic and turn it from a low tone to a high tone:

Code: Select all

ˈkảrw-a:n ‘cow-A.sg.’  >  ˈkảrw-a:n s-a:n ‘cow-A.sg. this-A.sg.’,  pra ˈkảrw-a:n s-a:n ‘for cow-A.sg. this-A.sg.’ 
  +   –                     +   –     +                              –   +         +

ˌgảlw-a:n ‘head-A.sg.’  >  gảlw-a:n ˈs-a:n ‘head-A.sg. this-A.sg.’,  pra gảlw-a:n ˈs-a:n ‘for head-A.sg. this-A.sg.’
  -   –                     –   –      +                               –  –   –      +           

The accent patterns this leads to are the same as with proclitics: with words belonging to the immobile accent paradigm (strong stem), enclitics are never accented, with words belonging to the mobile accent paradigm (weak stem), enclitics are unaccented if the suffix is strong, and accented if the suffix is weak. If there are both proclitics and enclitics, enclitics are accented in preference to proclitics.

This rule governing the behavior of enclitics is named Vasiľev-Dolobko’s law. It’s no longer active in any of the modern Slavic languages: in modern Slavic languages, enclitics all behave as weak morphemes; they are never accented because the two rules of accent assignment don’t allow this; proclitics, which also all behave as weak morphemes, can still be accented according to the two rules of accent assignment. However, Vasiľev-Dolobko’s law has left traces in all modern Slavic languages, most importantly the definite adjective declension and certain adverbs. Definite adjectives were formed from an indefinite adjective followed by an enclitic pronoun, and certain case forms of certain adjectives were liable to be accented on the enclitic according to Vasiľev-Dolobko’s law. Certain adverbs were formed from a noun followed by an enclitic pronoun, and certain case forms of certain nouns were again liable to be accented on the enclitic according to Vasiľev-Dolobko’s law. We’ll return to the accent of definite adjectives later when we’ll be dealing with diachrony.


4. Conclusion

So in the end, proclitics and enclitics are similar in some ways (they are only accented if the depend on an underlyingly unaccented word, never if they depend on an underlyingly accented word), and polar opposites in other ways (proclitics are accented with a word-initial low tone, enclitics are accented with a word-final high tone). And this apparently complex behavior can also be described with the two rules of accent assignment, provided we analyze all proclitics as weak morphemes and all enclitics as strong morphemes.

The ability of clitics to be accented is something unknown in Germanic and Romance languages (at least to me, to be honest), and while Greek has some pretty interesting phenomena concerning the accent of clitics they are very different from the ones Slavic has. Increasingly, however, even Slavic languages tend to make clitics unaccentable, or in more technical terms, to exclude them from the accentual domain (i.e. the two rules of accent assignment don’t apply to clitics, which are always unaccented). In my dialect of Croatian, clitics are unaccentable and there’s only one example of accent moving to a proclitic preserved as an adverb that I can think of. Other dialects, however, still accent proclitics according to the above rules (but have lost Vasiľev-Dolobko’s law and never accent enclitics). This is also a part of the standard language but is considered eccentric (or even hillbilly dialectal), except for verbal negation.

Once again I recommend everyone reading this thread to use the information found here to make a complex yet elegant (or less elegant, once we get to diachrony…) accent system for their conlang.

[Salmoneus]

This all seems quite straightforward... except for the idea of pre-accenting and post-accenting affixes. If you intend to write more, I'd be interested in hearing where they came from.

Pre-accenting is something that I know sometimes occurs in languages due to obscured resyllabification effects*.

But post-accenting seems difficult to produce through purely metric means, given that onsets don't normally determine weight.





*

Spoiler:

So, let's for example imagine a language with a C(C)V(C)(C) structure. Syllabification prefers CC onsets to cross-syllabic clusters. Open syllables are light, closed syllables are heavy.
So we could have roots:
dát
rást
(accent showing heavy syllable)

And the suffixes -któg, -kóg and -koga (through some great unfortunate coincidence!). And just for fun, the prefix ba-.

With the first root, that gives us the forms:
data - dát.któg - da.tkóg - da.tko.ga | ba.da.ta - ba.dát.któg - ba.da.tkóg - ba.da.tko.ga
And with the second root we get:
rást - rást.któg - rás.tkóg - rás.tko.ga| ba.rást - ba.rást.któg - ba.rás.tkóg - ba.rás.tko.ga

Now if we remove all final vowels, put stress on the first heavy syllable, or the first light syllable if none are heavy, drop coda fricatives, and drop the third C in a CCC cluster, we get the paradigms (now using accent to show stress):

mobile: dát - dátkog - datkóg - dátkog | bádat - badátkog - badatkóg - bádatkog
immobile: rát - rátkog - rátkog - rátkog | barát - barátkog - barátkog - barátkog

So in this system we quite simply produce mobile and immobile paradigms, and we have three suffixes -kog that differ only in how they affect mobile accent root: one is pre-accenting, one is self-accenting, and one is non-accenting.

But it's hard to see how a post-accenting affix could be produced by similar means!

[on stressed clitics - might German separable prefixes like vor- count? Unfortunately, 'clitic' can be a pretty vague concept!

[Zekoslav]

Most of pre-accenting and post-accenting suffixes in Slavic have arisen through conditioned accent shifts, towards the beginning of the word for pre-accenting and towards the end of the word for post-accenting suffixes. Others have arisen through whatever processes created the PIE. and Proto-Balto-Slavic accent systems... and every reconstruction of that is very speculative internal reconstruction: no phonological origin of the PIE. and PBSl. accent systems can be easily discerned. In theory it can be syllable weight (with vowel lenght, as in Romance, or without vowel lenght, as in your example) or tone (as in Japanese).

I'll write about it when I start dealing with diachrony. Other than sound changes, a lot of analogical levelling is involved.

[German, yes, that may be it. I'd need a closer look into it's history in order to be sure]

[gestaltist]

But it's hard to see how a post-accenting affix could be produced by similar means!

I'm nowhere close to an expert but being a native speaker of a Slavic language, my intuition is that this might be due to (some? all?) post-accenting morphemes losing their vowel when not final. I wonder what Zekoslav would have to say about that.