Pitch-accent language

A pitch-accent language is a type of language that, when spoken, has certain syllables in words or morphemes that are prominent, as indicated by a distinct contrasting pitch (linguistic tone) rather than by volume or length, as in some other languages like English. Pitch-accent languages also contrast with fully tonal languages like Vietnamese, Central Thai and Standard Chinese, in which practically every syllable can have an independent tone. Some scholars have claimed that the term "pitch accent" is not coherently defined and that pitch-accent languages are a sub-category of tonal languages in general.

Languages that have been described as pitch-accent languages include: most dialects of Serbo-Croatian, Slovene, Baltic languages, Ancient Greek, Vedic Sanskrit, Tlingit, Turkish, Japanese, Limburgish, Norwegian, Swedish, Western Basque, Yaqui, certain dialects of Korean, Shanghainese, and Livonian.

Pitch-accent languages tend to fall into two categories: those with a single pitch-contour (for example, high, or highlow) on the accented syllable, such as Tokyo Japanese, Western Basque, or Persian; and those in which more than one pitch-contour can occur on the accented syllable, such as Punjabi, Swedish, or Serbo-Croatian. In this latter kind, the accented syllable is also often stressed another way.

Some of the languages considered pitch-accent languages, in addition to accented words, also have accentless words (e.g., Japanese and Western Basque); in others all major words are accented (e.g., Blackfoot and Barasana).

The term "pitch accent" is also used to denote a different feature, namely the use of pitch when speaking to give selective prominence (accent) to a syllable or mora within a phrase.

Characteristics of pitch-accent languages

Definitions

Scholars give various definitions of a pitch-accent language. A typical definition is as follows: "Pitch-accent systems [are] systems in which one syllable is more prominent than the other syllables in the same word, a prominence that is achieved by means of pitch" (). That is to say, in a pitch-accent language, in order to indicate how a word is pronounced it is necessary, as with a stress-accent language, to mark only one syllable in a word as accented, not specify the tone of every syllable. This feature of having only one prominent syllable in a word or morpheme is known as culminativity.

Another property suggested for pitch-accent languages to distinguish them from stress languages is that "Pitch accent languages must satisfy the criterion of having invariant tonal contours on accented syllables ... This is not so for pure stress languages, where the tonal contours of stressed syllables can vary freely" (). Although this is true of many pitch-accent languages, there are others, such as the Franconian dialects, in which the contours vary, for example between declarative and interrogative sentences. This is not always true of pitch-accent languages, some of which, like Japanese and Northern Bizkaian Basque, have accentless words. But there are also some pitch-accent languages in which every word has an accent.

Characteristics of the accent

High vs. low accent

When one particular tone is marked in a language in contrast to unmarked syllables, it is usual for it to be a high tone. There are, however, a few languages in which the marked tone is a low tone, for example the Dogrib language of northwestern Canada, the Kansai dialect of Japanese, and certain Bantu languages of the Congo such as Ciluba and Ruund.

Disyllabic accents

One difference between a pitch accent and a stress accent is that it is not uncommon for a pitch accent to be realised over two syllables. Thus in Serbo-Croatian, the difference between a "rising" and a "falling" accent is observed only in the pitch of the syllable following the accent: the accent is said to be "rising" if the following syllable is as high as or higher than the accented syllable, but "falling" if it is lower (see Serbo-Croatian phonology#Pitch accent).

In Vedic Sanskrit, the ancient Indian grammarians described the accent as being a high pitch () followed by a falling tone () on the following syllable; but occasionally, when two syllables had merged, the high tone and the falling tone were combined on one syllable.

In Standard Swedish, the difference between accent 1 and accent 2 can only be heard in words of two or more syllables, since the tones take two syllables to be realised. In Värmland as well as Norrland accent 1 and 2 can be heard in monosyllabic words however. In the central Swedish dialect of Stockholm, citation forms of words with accent 1 have a LHL contour and a HLHL contour when marked with accent 2, with an additional peak in the second syllable.

In Welsh, in most words the accent is realised as a low tone on the penultimate syllable (which is also stressed) followed by a high tone on the final; but in some dialects this LH contour may take place entirely within the penultimate syllable.

Similarly in the Chichewa language of Malawi a tone on a final syllable often spreads backwards to the penultimate syllable, so that the word is actually pronounced Chich<u>ēwā</u> with two mid-tones, or Chichěwā, with a rising tone on the penultimate syllable. Sentence-finally it can become Chich<u>ěwà</u> with a rising tone on the penultimate and a low tone on the final.

Peak delay

A phenomenon observed in a number of languages, both fully tonal ones and those with pitch-accent systems, is peak delay. In this, the high point (peak) of a high tone does not synchronise exactly with the syllable itself, but is reached at the beginning of the following syllable, giving the impression that the high tone has spread over two syllables. The Vedic Sanskrit accent described above has been interpreted as an example of peak delay. Ancient Greek "it demands".

Forwards spreading

Forwards spreading of a tone is also common in some languages. For example, in the Northern Ndebele language of Zimbabwe, the tonal accent on the prefix ú- spreads forward to all the syllables in the word except the last two: "to laugh"; "to make one another laugh". Sometimes the sequence HHHH then becomes LLLH, so that in the related language Zulu, the equivalent of these words is and with an accent shifted to the antepenultimate syllable. Because the accent is both stressed and high-pitched, Persian can be considered intermediate between a pitch-accent language and a stress-accent language.

More complex pitch accents

In some simple pitch-accent languages, such as Ancient Greek, the accent on a long vowel or diphthong could be on either half of the vowel, making a contrast possible between a rising accent and a falling one; compare () "at home" vs. () "houses". Similarly in Luganda, in bimoraic syllables a contrast is possible between a level and falling accent: "Buganda (region)", vs. "Baganda (people)". However, such contrasts are not common or systematic in these languages.

In more complex types of pitch-accent languages, although there is still only one accent per word, there is a systematic contrast of more than one pitch-contour on the accented syllable, for example, H vs. HL in the Colombian language Barasana,

Because of the number of ways languages can use tone, some linguists, such as the tonal languages specialist Larry Hyman, argue that the category "pitch-accent language" can have no coherent definition, and that all such languages should simply be referred to as "tonal languages".

The ancient Indian grammarians describe the accented syllable as being "raised" (), and it appears that it was followed in the following syllable by a downwards glide, which the grammarians refer to as "sounded" (). In some cases, language change merged an accented syllable with a following svarita syllable, and the two were combined in a single syllable, known as "independent svarita".

The precise descriptions of ancient Indian grammarians imply that the was characterised by rising pitch and the by falling pitch. In the tradition represented by the Rigveda, a collection of hymns, the highest point of the accent appears not to have been reached until the beginning of the syllable. In other words, it was an example of "peak delay" (see above).]]

Baltic tones are often classified as either "acute" or "circumflex". However, these labels indicate a diachronic correspondence rather than a phonetic one. For example, the "acute" accent is falling in Lithuanian but a high level tone in Latvian and is presumed to have been rising in Old Prussian and Classical Greek. The "circumflex" is rising in Lithuanian but falling in Latvian, Prussian and Classical Greek.

In the tree diagram on the right, as adopted from Poljakov, names for (original) Baltic tones have been equated with those of modern Standard Lithuanian and the falling tone in Latvian is depicted as derived from a Baltic rising tone. According to some it was Lithuanian that "switched the places" of the Baltic tones.

Lithuanian

Long segments in Lithuanian can take one of two accents: rising or falling. "Long segments" are defined as either long vowels, diphthongs or a sequence of a vowel followed by a sonorant if they are in a stressed position. Pitch can serve as the only distinguishing characteristic for minimal pairs that are otherwise orthographically identical, e.g., 'time:gen.pl' vs. 'hang:irr.3' (rising and falling tone indicated by a tilde and an acute accent respectively.)

Latvian

In Latvian, long segments (the same criteria as in Lithuanian) can take on one of three pitches ( or more specifically ) either ("level"), ("broken") or ("falling") indicated by Latvian linguists with a tilde, circumflex or a grave accent respectively (in IPA, however, the tilde is replaced by a macron because the former is already reserved to denote nasalized vowels.) Some authors note that the level pitch is realized simply as "ultra long" (or overlong.) Although the (Indo-European) Latvian and (Uralic) Livonian are phylogenetically unrelated (being from different language families) both have influenced each other heavily in terms of phonology. Whether Livonian acquired this feature from Latvian or vice versa is debated; however, as Livonian is the only Finno-Ugric language to have this feature, the majority of researchers believe it was a product of Latvian influence on Livonian and not the other way around.]]

A pitch accent is found in the following Rhinelandic languages or dialects: Limburgish, Ripuarian and Moselle Franconian (excluding Luxembourgish). They are sometimes collectively referred to as West Central German tonal languages.

In these dialects there is a distinction between two different tonal contours, known as "tonal accent 1" and "tonal accent 2". As with Lithuanian, the distinction is made only in stressed syllables and, for the most part, only when the syllable contains a long vowel or diphthong or vowel that is followed in the same syllable by a sonorant (r, l, m, n, ŋ). No distinction of tones is made in stressed syllables containing a short vowel only. Although the accentual system resembles that of Swedish, the two are thought to have arisen independently. Unlike Swedish, where the distinction in tones is not made in monosyllables (except for in northern and western dialects), in the Franconian dialects it very frequently occurs in monosyllables, e.g., (Ripuarian dialect) "sieve" vs. "she".

The two accents have different realisations in different dialects. For example, in Cologne, accent 1 has a sharp fall near the beginning of the syllable, and accent 2 remains level for a while before falling. In Arzbach, near Koblenz, on the other hand, accent 1 rises slightly or remains level, while it is accent 2 that falls sharply, that is, more or less the reverse of the Cologne pattern. In Hasselt in Belgian Limburg, accent 1 rises then falls, and with accent 2 there is a fall and then a rise and another fall. The three types are known as Rule A, Rule B and Rule 0, respectively.

Serbo-Croatian

The Neoštokavian dialect used for the basis of standard Bosnian, Croatian and Serbian distinguishes four types of pitch accents: short falling (ȅ), short rising (è), long falling (ȇ), and long rising (é). There are also unaccented vowels: long (ē) and short (e). The accent is said to be relatively free, as it can be manifested on any syllable except the last. The long accents are realized by pitch change within the long vowel; the short ones are realized by the pitch difference from the subsequent syllable.

Accent alternations are very frequent in inflectional paradigms by both types of accent and placement in the word (the so-called "mobile paradigms", which were present in the PIE itself but became much more widespread in Proto-Balto-Slavic). Different inflected forms of the same lexeme can exhibit all four accents: 'pot' (nominative sg.), (genitive sg.), (nominative pl.), (genitive pl.).

Restrictions on the distribution of the accent depend on the position of the syllable but also on its quality, as not every kind of accent is manifested in every syllable.

A falling tone generally occurs in monosyllabic words or the first syllable of a word ( 'belt', 'horn'; 'old woman', 'river ship'; 'small house', Kȃrlovac). The only exception to this rule are the interjections, i.e., words uttered in the state of excitement (, )
A rising tone generally occurs in any syllable of a word except the ultimate and never in monosyllabic words ( 'water', 'harbour'; 'meadow', 'slam'; 'female orphan', 'beginning'; 'wormhole', 'liberation').

Thus, monosyllables generally have falling tone, and polysyllabic words generally have falling or rising tone on the first syllable and rising in all the other syllables except the last. The tonal opposition rising vs. falling is generally possible only in the first accented syllable of polysyllabic words, but the opposition by length, long vs. short, is possible even in the nonaccented syllable and the post-accented syllable (but not in the preaccented position).

Proclitics (clitics that latch on to a following word), on the other hand, can "steal" a falling tone (but not a rising tone) from the following monosyllabic or disyllabic words (as seen in the examples ). The stolen accent is always short and may end up being either falling or rising on the proclitic. That phenomenon is obligatory in Neoštokavian idiom and therefore in all three standard languages, but it is often lost in spoken dialects because of the influence of other dialects (such as in Zagreb because of the influence of Kajkavian dialect).

{| class=wikitable

!colspan=2|in isolation||colspan=2|with proclitic

|rowspan=2|rising

|||I want

|||I don't want

|||inability

|||not being able to

|rowspan=4|falling

|N: , A: ||winter

| (A)||in the winter

|||I see

|||I can't see

|N, A: ||city

| (A)

|to the city (stays falling)

|N: ||forest

| (L)

|in the forest (becomes rising)

Slovenian

In Slovenian, there are two concurrent standard accentual systems: the older, tonal, with three "pitch accents", and the younger, dynamic (i.e., stress-based), with louder and longer syllables. The stress-based system was introduced because two thirds of Slovenia has lost its tonal accent. In practice, however, even the stress-based accentual system is just an abstract ideal, and speakers generally retain their own dialect even when they try to speak Standard Slovenian. For example, speakers of urban dialects in western Slovenia without distinctive length fail to introduce a quantitative opposition when they speak the standard language.

The older accentual system is tonal and free ( 'strawberry', 'raspberry', 'master, lord'). There are three kinds of accents: short falling (è), long falling (ȇ) and long rising (é). Non-final syllables always have long accents: 'crustacea', 'sinew'. The short falling accent is always in the final syllable: 'brother'. Three-way opposition among accents can only then be present: 'board' : 'goods, ware' : 'lady'. Accent can be mobile throughout the inflectional paradigm: — , — , — , , — ). The distinction is made between open and (either long or short) and closed and (always long).

Basque

The Basque language of northeastern Spain and southwestern France has a number of different dialects and a number of accentual patterns. Only western varieties seem to have a tonal accent, and eastern varieties have a stress accent (the stress-accent dialects also differ one from another). Northern Bizkaian has most nouns accentless in their absolutive singular form, but they have a default high tone (shown by underlining below), which continues throughout the word except for the first syllable. These examples come from the Guernica (Gernika) dialect:

"Guernica"
"forest"
"potato"
"kidney"

There are, however, a few nouns (often borrowings) with a lexical accent. As in Japanese, the accent consists of a high tone, followed by a low one:

"Bilbao"
"supper"

In addition, some suffixes (including all plural suffixes) are preaccenting and so cause an accent on the syllable before the suffix:

"women"
"of dogs" (genitive plural)
"from Guernica"

Other suffixes do not cause any extra accent:

"of the dog" (genitive singular)

When a preaccenting suffix is added to an already-accented word, only the first accent is retained:

"from Bilbao"

The accent from Ondarroa is similar but the accent of the word, if any, always appears on the penultimate syllable:

"from Bilbao" (Ondarroa pronunciation)

Intonation studies show that when an accentless word is spoken either in isolation or before a verb, it acquires an accent on its last syllable (or, in Ondarroa, on its penultimate syllable). However, that is an intonational accent, rather than a lexical accent:

"the friend () has come"

When an accentless word in those dialects of Basque is followed by an accented word, the automatic high tones continue in a plateau as far as the accent:

"John's friend's grandmother"

Another pitch accent area in Basque is found in western Navarre, near the border with France in the towns of Goizueta and Leitza. There is a strong stress accent there on the second or the first syllable of every word, like with central dialects of Basque, but there is also a pitch contrast superimposed on the stress: (rise-dip-rise) "the mountain" vs. (rise-fall) "the mountains".

Turkish

Although the Turkish accent is traditionally referred to as "stress", recent studies have pointed out that the main correlate of lexical accent is actually pitch. In a word like "with a word", the accented second syllable is thus higher than the other two but has less intensity (loudness).

The accented syllable is slightly higher in pitch than the following syllable. All other words, when pronounced in isolation, either have a slightly raised pitch on the final syllable or are pronounced with all the syllables level. Others, such as , prefer the traditional view that the final accent in such words is a kind of stress.

Persian

The accent of Persian words used to be always referred to as "stress" but is recognised as a pitch accent in recent works. Acoustic studies show that accented syllables have some of the characteristics of stressed syllables in stress-accent languages (slightly more intensity, more length, more open vowels), but that effect is much less than would normally be expected in stress-accent languages. The main difference is one of pitch, with a contour of (L)+H*.

{| class="wikitable"

|+Examples

!Hangul

!IPA

!English

|daughter-in-law

|mother

|native speaker

|elder brother

Shanghainese

The Shanghai dialect of Wu Chinese is marginally tonal, with characteristics of a pitch accent.

Not counting closed syllables (those with a final glottal stop), a monosyllabic Shanghainese may carry one of three tones: high, mid, low. The tones have a contour in isolation, but for the following purposes, it can be ignored. However, low tone always occurs after voiced consonants and only then. Thus, the only tonal distinction is after voiceless consonants and in vowel-initial syllables, and there is only a two-way distinction between high tone and mid tone.

In a polysyllabic word, the tone of the first syllable determines the tone of the entire word. If the first tone is high, the following syllables are mid. If it is either mid or low and not closed, the second syllable is high, and any following syllables are mid. If it is low and closed, the following syllables are mid. Thus, a mark for the high tone is all that is needed to note the tone in Shanghainese:

{| class=wikitable

!Romanization

!Hanzi

!Pitch pattern

!English

|No voiced initial (mid tone)

|mid–high–mid–mid

|Australia

|No voiced initial (high tone)

|high–mid–mid–mid

|bus

|Voiced initial (not closed)

|low–high–mid

|Shanghainese person

|Voiced initial (closed)

|low–mid–mid

|Japanese person

Bantu languages

The Bantu languages are a large group of about 550 languages, spread over most of southern and central Africa. Proto-Bantu is believed to have had two tones: H and L. It does not appear to have had a pitch-accent system, as defined above, since words with such forms as HL, HH, LH, and LL were all found: *káda "charcoal", *cómbá "fish", *nyangá "horn" and *tope "mud". In other words, words like *cómbá could have two high tones, while others had one tone or none.

However, over time, processes such as Meeussen's rule, by which sequences such as HHH became HLL, LHL, or LLH, tended to eliminate all but one tone in a word in many Bantu languages, making them more accent-like. Thus in Chichewa, the word for "fish" () now has HL tones, exactly like the word for "charcoal" ().

Another process that makes for culminativity in some Bantu languages is the interaction between stress and tone. The penultimate syllable of a word is stressed in many Bantu languages, and some of them tend to have high tones on the penultimate syllable. For example, in Chitumbuka, every phonological phrase is accented with a falling tone on the penultimate: "we are cooking porridge". In other languages, such as Xhosa, the high tone is attracted to the antepenultimate although the penultimate being stressed.

Ciluba and Ruund, in the Democratic Republic of the Congo, are two Bantu languages that are interesting for their "tone reversal". Low tone is phonologically active in places that other languages of the same family have a high tone. Thus, in a word like *mukíla "tail", most other Bantu languages have a high tone on the second syllable, but Chiluba has and Ruund has , with a low-toned accent.

Luganda

Luganda, a language of Uganda, has some words with apparent tonal accents. They can be either high or falling (rising tones do not occur in Luganda). Falling tones are found on bimoraic syllables or word-finally:

"country"
"city"
"hospital"

Some words, however, have two accents, which are joined in a plateau:

"Kampala"

Other words are accentless:

"book"

However, accentless words are not always without tones but usually receive a default tone on all syllables except the first one or the first mora:

"book"
"Bunyoro" (name of region)

A double consonant at the beginning of a word counts as a mora. In such words, the first syllable also can have a default tone:

"Toro" (a region)

Default tones are also heard on the end of accented words if there is a gap of at least one mora after the accent (the default tones are lower in pitch than the preceding accent):

"south"
"hospital"

The default tones are not always heard but disappear in certain contexts, such as if a noun is the subject of a sentence or used before a numeral:

"Mbarara is a city"
"ten books"

In some contexts, such as affirmative verb + location, or phrases with "of"), the high tone of an accent (or of a default tone) can continue in a plateau until the next accented syllable:

"in the south of Uganda"
"he is in Buganda"

However, the situation with verbs is more complicated since some verbal roots have inherent word accent, and the prefixes added to the verb often have an accent. Also, some tenses (such as negative and relative clause tenses) add an accent to the final syllable.

When two or three accents appear in a row in a verb, H-H becomes H-L, and H-H-H becomes H-L-L. However, the default tones are not added on the syllables with deleted accents, which leads to forms like (from ) "they will see". There, not one but two low-toned syllables follow the accent.

South African English

In Broad South African English, (phonetically ) is often deleted, such as in word-initial stressed syllables (as in house), but at least as often, it is pronounced even if it seems to be deleted. The vowel that follows in the word-initial syllable often carries a low or low rising tone. In rapid speech, that can be the only trace of the deleted . Potentially minimal tonal pairs are thus created, like oh (neutral or high falling ) vs. hoe (low or low rising ).

Welsh

In Welsh a stress accent usually comes on the penultimate syllable (with a few exceptions accented on the final, such as the word "Welsh"), and is usually on a low pitch followed by a rising pitch. "In Welsh, the stressed syllable is associated with lower pitch than less stressed or unstressed syllables ... However, the post-stress syllable in Welsh is typically produced on a higher pitch." It is believed that this came about because late Brythonic (the ancestor of Welsh) had a penultimate accent that was pronounced with a high pitch. When the final vowels of words were lost, the high pitch remained on what was now the final syllable, but the stress moved to the new penultimate. Thus LHL changed to LH, with the stress on the low syllable.

Although it is usually said that the high pitch is in the final syllable of the word, an acoustic study of Anglesey Welsh found that in that dialect at least the peak of the tone was actually in the penultimate syllable, thus the last two syllables were L+H* L.

Scottish Gaelic

Several northern dialects of Scottish Gaelic, such as those of Lewis, Wester Ross and Strathnaver, feature pitch accents which derive from vowel hiatus. For instance, the words "cow" and "reef" are a minimal pair, but the former is pronounced with a level or rising tone, as , and the latter with a falling tone beginning rising or high, as . A similar tonal contrast is found in ostensibly disyllabic words. Scottish Gaelic has a phenomenon of epenthesis whereby certain consonant clusters (see here for a full list) feature a repeating, equally stressed vowel, creating a second syllable whose vowel is considered a unit with, or echo of, the first. This can be seen in the minimal pair "to leave", pronounced with a rising tone across the word and both syllables equally stressed, and "empty", pronounced with a rising tone followed by a falling one.

Ternes claimed that Scottish Gaelic is inherently a pitch-accent language, which manifests in the dialects of Argyll as glottalisation akin to stød.

Yaqui

The Yaqui are a native American people living mostly in Mexico but also in Arizona. About 17,000 people are said to speak Yaqui, which is a Uto-Aztecan language.

Yaqui has a tonal accent in which the accent is on the first or the second mora of the word. A long vowel has two moras in Yaqui, and a short vowel or diphthong has one mora. After the accent, the high tone continues with a very slight decline until the end of the word.

About two thirds of words have an accent on the first mora, and all tones of the word are then high: