A Change in Welsh from
|
| canwyllerni | /kan uï λer ni/ | /2 3 1 2/ | 'candlesticks' |
| caniadau | /kan ja de-/ | /3 1 2/ | 'songs' |
| caniad | /kan jad/ | /1 2/ | 'song' |
| can | /kan/ | /1/ | 'white' |
There are various phonological rules that are applied through stress. Perhaps the most obvious is an intrusion of aspiration in certain syllables when the addition of a suffix shifts the syllable from ultimate to penultimate position. For example, brenin 'king' has such a syllable in the ultimate position; and when the feminine ending -es is added to it, we obtain the word brenhines 'queen'. The phenomenon does not occur generally throughout the language, however, but is limited to particular nouns and verbs. While these words do tend to share certain phonological characteristics, they are not entirely predictable. Indeed, the only way to predict them phonologically is to consider the process to be a loss of aspiration in the ultima (a solution which runs into some difficulty, however, when it interacts with the contract syllable below).
The pitch accent is characterised by a steady rise in pitch (the fundamental frequency) through the word, culminating in the high pitch accent in the ultima. Traditionally, this pattern has been regarded as the ancillary accent (see, for example, Watkins 1961:29), partly as the result of an assumption on the part of linguists working in English and other languages and partly due to terminology borrowed from Latin by the medieval bardic grammarians and used in the Latin sense (Welsh acen 'accent' < Latin accentus).
The foremost phonological rule motivated by the pitch accent is a vowel centralisation rule in which the high vowel is lowered to the mid central position when the syllable in which it is realised shifts out of the pitch accented ultima. The letter y is pronounced as the high central vowel [ï] or the high front vowel [i] (depending upon dialect) in the accented ultima, but it is reduced to the mid central vowel [ə] in other positions (including proclitic position, where it is not realised with pitch accent). Thus, for example, the word mynydd [mənïð] 'mountain' is realised in the plural as mynyddoedd [mənəðoïð].
Likewise, when the high front vowel w [u] shifts out of the pitch accented ultima, it is reduced to the mid central vowel y [ə]. For example, the word bwrdd [burð] 'table' is realised in the plural as byrddau [bərðε] (or [bərða], once again depending upon the dialect).
The rule of vowel centralisation (which is actually much more extensive than this - see Morris Jones 1913:116-20) is phonologically motivated and is not dependent upon morphology or lexicon. It is by this and other such phonological rules that we can determine that the principal accent in Welsh is not stress, but pitch.
The status of pitch as principal and stress as ancillary has evolved from several developments going back to Old Welsh (from about 800 to about 1175). There is a considerable amount of controversy surrounding the accent in Old Welsh, with some linguists arguing that the ultima had a pitch accent only (for example Watkins 1972), and others claiming that the ultima also maintained some degree of stress (for example, Jackson 1975/76). These arguments are summarised in Griffen 1991/92 (as they affect the development of epenthesis).
Whether there was a stress accent on the ultima in Old Welsh or not, in Middle Welsh the primary stress accent either shifted to or developed in the penult, where it has remained. This shift or development, however, came after the establishment of the major phonological rules of Welsh, and the principal accent pattern has thus remained the pitch accent pattern. Nonetheless, as we shall see below, there have been some rather surprising developments in Modern Welsh (since about 1500).
Pitch accent in cynghanedd poetry
Perhaps the most obvious case for the primacy of pitch accent over stress accent in Welsh is found in the poetic rules of cynghanedd ('alliteration, harmony' - see Morris Jones 1925, Parry-Williams 1936, Rowlands 1976). This highly intricate system of correspondence involving consonants, vowels, scansion, and rhyme is always based upon the oral pronunciation of the language, and never upon the writing (compare especially Griffen 1981).
In the scansion/rhyme scheme of the cynghanedd, a monosyllabic word must by convention often rhyme with a word of two or more syllables. When this occurs, the monosyllable which combines both the high pitch and the heavy stress accent must rhyme with an ultima which has the high pitch accent but not the heavy stress accent. For example, the following rhyme is cited by Morris Jones (1925:258 - my translation):
Cywir yw'ch gweirwir gárwr,
Cywir ar y gwir yw'r gr.
Loyal is your truthful kinsman,
Loyal to the truth is the man.
We can see the effect of this type of scansion and rhyme in the English adaptation of cynghanedd in Lloyd 1985. Lloyd (1985:5) suggests such rhymes as the following:
Night may dare not my dearest
Shadow throw where she doth rest.
The scansion/rhyme scheme cited here obviously does not work for English, because pitch and stress accent in English always co-occur in one syllable. In Welsh, on the other hand, the scheme does work, because the syllables being rhymed in Welsh always maintain the pitch accent. It is thus clearly the pitch accent that is the principal, with the stress accent ancillary.
Once again, the primacy of pitch accent reflects the situation in the language and poetry of Welsh in practice, not in linguistic or poetic description. While the rules of cynghanedd use the term acen 'accent' to refer to the stressed penult, we must bear in mind that this term is borrowed from Latin and is greatly influenced by English and that it is used with reference to the phonological systems of these other languages. Had Wales been contiguous with France (with its rather transparent Celtic substratum), the descriptive tradition would probably have been more accurate.
Welsh penillion singing
With the heavy stress accent occurring in effect on a phonologically 'unaccented' syllable, we may wonder how one sings the language with a musical instrument that emphasises the rhythm through a 'stress' beat - a beat characterised by a patterned rise in amplitude. We might expect the musical beat to coincide with the linguistic stress since the two phenomena share like characteristics in amplitude; but this would be inappropriate to the linguistic patterning of the language, in which the marked emphasis should co-occur with the pitch accented ultima. This, however, would require the beat to be marked not by heavy amplitude, but by high pitch - the music would be governed by a high note on the 'beat', eliminating melody independent of rhythm.
We are thus faced with two horns of a dilemma. Either the singer pronounces the language unnaturally in order to adhere to the rhythmic beat of the music, or the musician plays an unnatural rhythm governed by a higher note corresponding with every pitch accented syllable (or at least those involved in rhyme). Welsh singers and musicians have found a unique way to pass between the horns of this dilemma in what is known as penillion singing.
As intricate as the correspondence rules of cynghanedd are the rules that join the poem with harp music in penillion ('verses, stanzas' - see especially Davies 1983). As can be seen in the Ap Huw Manuscript (Polin 1982), the traditional medieval harp accompaniment is highly repetitive, often repeated bar by bar with a slight culminative change at the end (see also Kinney 1979). In more modern penillion, however, the harp often plays an intricate tune or air. The language of the poetry follows a different melody altogether, allowing the poet to sing or declaim to a melody not connected with the rhythm of the instrument. The disconnection between rhythm and accent is related by Polin: "A London eisteddfodd [sic] was described as an occasion on which a blind minstrel on the triple harp accompanied penillion sung 'by 3 singers, each on a different accent'" (1982:34).
It is significant that penillion singing, while allowing the linguistic accent pattern to coexist with the musical rhythm, has established a protocol in which the voice follows the harp, and the vocalist is responsible for accommodating the poem to the music. Thus, R. Roberts (1914:150) observes:
...the harp does not accompany the voice, but rather the reverse takes place, i.e., the voice accompanies the harp. It is the instrument that leads, and it plays the tune from beginning to end, regardless and independently of the singer; its duty being to give a steady, well-timed, and correct rendering of the melody from the first to the last note. The voice has to 'strike' or tarro on the proper beat or fraction of a beat, as the length of the metre (mydr) may demand, and to cantillate concurrently with the melody, but to contrive so as to end simultaneously with the harp. If the singer fails to accomplish this, he puts himself 'out of court' (allan o wellt)...
Not only do the melodies of the vocalist and musician differ, but even the meter or rhythm can be different. It is quite typical for the vocalist to sing in triple meter (for example, 3/4 time) while the harper plays a tune in duple meter (for example, 2/4 or 4/4 time). Nor is the vocalist permitted to slur words or to extend them unnaturally over a series of notes so that an accent might seek out a beat.
Penillion singing therefore allows both the vocalist and the musician to ply their skills within their own frameworks. The language is natural, with the principal accent pattern realised in the culminative high pitch of the ultima. The music is also natural, with the rhythm realised in proper time and the beat characterised by amplitude, not by pitch.
There is one aspect of the penillion pattern, however, that has doubtless contributed to the linguistic change noted below: The onus is on the vocalist to adapt the sung melody to the rhythm of the harp. While this cannot cause unnatural articulations, it does require a constant awareness of the musical rhythm so that the voice may accompany the instrument.
The advent of European rhythmic song
While it is still widely performed in Wales along with other types of song, penillion singing was the only form of song before the Early Modern period - at least the only form that we may (from a modern perspective) describe as an "art song." Indeed, as R.E. Roberts notes, "prior to Tudur Aled [ca. 1465-ca. 1525] there was no such pursuit as composing music for words, or words for music. One made poetry, another made music, and a third united the two and sang the wedded poetry and music" (1925/26:20).
In the Early Modern period though, rhythmic European "art" music was introduced under the Tudor dynasty at a time when the dynamic new 'isorhythmic' music (or polyphony) was sweeping Europe (Strohm 1993). The new approach to rhythm brought fundamental changes to Welsh music, as it emphasised the beat from the instrument and required it to coincide not with the pitch accent, but with the heavy stress of the penultimate syllable.
The advent of this new European rhythmic singing was marked by an immediate reaction in the highly formal metrical form of cynghanedd known as the cywydd. Within a line (but never between lines) there could be an internal rhyme of a monosyllable with the stressed penult - a stress-based rhyme. By the time of Tudur Aled (noted above), however, this form all but disappeared as another form that emphasised the intricacy of consonant alliterations gained in prominence (see Rowlands 1976:xlviii). Such a development would certainly have been a distancing of the poetic establishment from the new emphasis on the stress accent and a reasserting of traditional consonant correspondences unconnected with stressed syllables.
There was a later attempt to apply the rhythmic music to penillion singing. In the mid seventeenth and the eighteenth centuries, such poets as Huw Morus composed a new 'accentuated' cynghanedd, in which "the musical phrase was always very strictly observed and the words made to conform to the framework of the tune" (Parry 1962:552; compare B.J. Williams 1989:227). The experiment, however, did not prevail, and the cynghanedd with its penillion singing returned to its tradition and developed separately from rhythmic song, which may also have experienced an extensive infusion from English (compare J. Ll. Williams 1909).
Perhaps the most widely known area of the new Welsh rhythmic song has been in church hymns such as Hyfrydol, Cwm Rhondda, and many other tunes that have found their way into English hymnals. Of course, such religious music became very popular in Wales at this very time due to the Reformation and subsequent developments (see Parry Williams 1948). Other, less widely known songs following the European rhythmic model include the popular Welsh folk song and ballad, the stable loft song (Saer 1980), and the plygain carol (Saer 1977 - compare also Gwynn Williams 1971).
These new kinds of songs returned the singer and the musician to the horns of the dilemma noted above - either the language must conform to the music, or the music must conform to the language. Since the music was now an international European "art" song that overwhelmed local traditions and since the vocalist was already used to following the instrument anyway, it was the language that had to change.
Traditional pitch-motivated deletion
The linguistic change brought about by the change in music can be seen in rules of syllable deletion. Traditionally, syllable deletion has affected the initial syllable of a disyllabic word. Thus, for example, Jones notes that "in some cases in normal colloquial use the penultimate is actually lost; this seems to occur in a few disyllables which are weakly stressed as a whole in the sentence-context and in which the vowel of the final syllable is of greater natural sonority than that of the initial syllable: 'to for eto 'yet', 'ma, 'na, 'cw ('co) for yma, yna, acw 'here', 'there', 'yonder', 'na for dyna 'there is', 'voilà'. No doubt the frequently higher pitch of the final syllable plays a part in this process" (1949:63).
This pattern of syllable deletion very clearly indicates that the syllable with pitch accent is more prominent than the syllable with stress accent. In each case, the stress accented syllable is deleted and the pitch accented syllable is maintained. In fact, the environmental motivation for the deletion is the pitch accent, and the syllable being deleted is marked for the heavy stress of the primary stress accent (a point to which we return below).
Innovative stress-motivated syllable deletion
As discussed at length elsewhere (Griffen 1979; see also Morris Jones 1913:54), however, there is a newer pattern of syllable deletion, one based upon the stress accent pattern and independent of pitch. In this form of syllable deletion, the syllable with the weak tertiary phonological stress is dropped. For example, a word such as tragywyddol [tra gə: uï ðo-l] 'eternal' with a phonological /2 3 1 2/ stress pattern can be realised as [tra guï ðo-l] with a /3 1 2/ stress pattern.
This new pattern of syllable deletion simply does not fit into the overall patterning of the Welsh language, at least from a formal structural standpoint. For example, in the traditional generative model of phonology, other processes involving stress assignment apply at the underlying level only, while this new rule of syllable deletion applies at the surface. The interaction of these rules thus necessitates two separate applications of stress pattern assignment, one at the underlying level and one at the surface level (compare Griffen 1977 - see below).
The new rule of syllable deletion is contrary to the phonological system of Welsh. Far from being unsystematic, however, it constitutes what one might call an 'antisystematic' pattern. This is more than what we traditionally term a minor rule - it is a pattern that contradicts the overall functioning of the phonological structure. What could have caused such an inherently antisystematic development in the Welsh language?
Musical motivation for the innovative rule
The answer to this riddle lies in the date of the change. According to Morris Jones (1913:54), such 'shortened forms' occur only rarely in Middle Welsh (and then perhaps only in late Middle Welsh). Considering the apparent universality of the deletion of the most weakly stressed syllable (apparently universal due to the fact that stress seldom fails to coincide with pitch), its rarity in Middle Welsh is somewhat remarkable. Indeed, the phenomenon does not become widespread until the Early Modern Welsh of the sixteenth century.
The ascendancy of the new antisystematic rule, then, corresponds precisely with the advent of European "art" song brought to Wales under the Tudors. It is suggested here that the change in the language was caused by the requirement in song for rhythm emphasised by musical instrumentation to correspond with the stress accent without regard to the linguistically principal pitch accent. This new emphasis on the stress accent and its necessary cooccurrence with rhythmic beat (musical stress or amplitude) apparently developed speech habits in which the speakers became more aware both of the emphasis placed on the penult and of the lack of stress on the antepenult. Evidently, this process is still in progress, for as B.J. Williams points out, "Another phonetic phenomenon adding prominence to the penult is rhythm, or rather the tendency towards it" (1989:210).
Indeed, at this very same point in history, another deletion took place - one that was rare but quite indicative of what was going on in the language. In a few disyllables, the final, pitch-accented syllable deleted, yielding, for example, ond [ond] 'but' from onid, mynd [mïnd] 'to go' from myned [mə:nεd], and tyrd [tïrd] 'come!' from tyred [tə:rεd] (Morris Jones 1913:55). Clearly, such a development could not have occurred unless at least some influential speakers of the language were viewing the stress accent pattern as the principal; although to be sure, the traditional rule in which eto was realised as 'to was still very much in effect.
In the traditional rule of syllable deletion, it was the syllable before the primary pitch accent that was deleted, in spite of the fact that the deleted syllable happened to maintain the primary stress. The major new rule has come to be applied in basically the same manner as the old rule, except that now the syllable being deleted comes before the primary stress accent, the accent that is emphasised in rhythmic "art" song. Both rules occur simultaneously in the language.
Importance of music in the Welsh culture
How could song be so influential in the development of a language? Indeed, both song and poetry are of central importance to the Welsh culture. Both are presented in balanced measure in the local, national, and even international song and poetry festivals known in Welsh as the eisteddfodau, a term borrowed into the English language as the eisteddfods.
The fact is that song is so important to the Welsh culture that a change in the structure of the song can quite conceivably bring about a change in the structure of the language. On the other hand, poetry is also such an important force in the culture and its language that the extent of the linguistic change is in effect limited to the point that it appears to be antisystematic.
As such linguistic change originates or at least spreads from speakers of high prestige, we should note that these cultural priorities give both vocalists and poets an extremely influential position in the speech community. The manners in which the singers sing and the poets recite are thus subject to a great deal of imitation. Insofar as song is concerned, the popularity of hymn singing not only in church, but also in the widespread hymn festival or y gymanfa ganu has further ensured the spread of the rhythmic patterns established in the European model and emphasising the stress accent pattern.
Thus, we see that a change in music may indeed bring about a linguistic change of a rather drastic degree. While the principal accent pattern of Welsh language and literature has traditionally been the pitch accent, the requirements of European rhythmic song have brought about a change in the structure of Welsh phonology, with a new emphasis on the stress accent pattern, creating a situation in which the stress accent system in some phonological rules takes precedence over the pitch accent system. Meanwhile, however, the traditional pitch accent system still maintains precedence in its set of phonological rules.
Implications for other languages
Although we might wish to consider the situation in Welsh to be an idiosyncrasy of a culture that assigns extreme prestige to vocalists, musicians, and poets, we ought not lose sight of the fact that the Welsh situation is not that idiosyncratic. At present, popular Western music is making demands on tone languages such as Chinese that require the vocalist to suppress proper tone for the sake of the melody and rhythm. Moreover, the heavy beat of Anglophone popular music must also coincide with the phonological systems of such pitch accented languages as French, which maintains a system of accent highly reminiscent of Old Welsh and a tradition of song declamation as well. It would thus be of interest for linguists in such languages to observe the effects of popular Western music on the pronunciation of song lyrics and to anticipate the changes that might occur in the languages themselves.
Implications for linguistic theory
Finally, we should consider the implications of this development on phonological description and theory. As noted above, the antisystematic nature of the new rule of syllable deletion has already caused problems for the generative model. Two decades ago, this researcher reported these problems to the Linguistic Society of America (Griffen 1977). More recently, these issues have been addressed in the context of song and accent discrepancy in the Linguistic Association of Canada and the United States (Griffen 1995).
As we have seen above, the word tragywyddol [tra gə: uï ðo-l] 'eternal' with a phonological /2 3 1 2/ stress pattern can be realised as [tra guï ðo-l] with a /3 1 2/ stress pattern, showing that the stress assignment must be applied at the surface level. On the other hand, yet another syllable deletion rule - known as the 'contract syllable' - requires stress assignment on the underlying level only. For example, the underlying string /par/ (root) + /ha/ (factitive suffix) + /ï/ (verb-noun suffix) must be assigned a /3 1 2/ stress pattern which is then realised through the contraction of the penult and ultima as the disyllabic parhau /par haï/ 'to continue' with a /3 1/ stress pattern - not the /1 2/ pattern that would result from a surface application of stress assignment.
If the stress assignment occurs at the underlying level for the contract syllable, then it must apply at the underlying level for the phonology in general. But in the contract syllable, it does not reapply at the surface level; while in tertiary stress deletion, it must reapply. At the time of the report, the problem was seen as a violation of the noniterative rule convention, which was necessary in order to prohibit rules from applying to their own output and thus from reducing the entire lexicon to disyllables or even to monosyllables (in the case of the contract verbs).
Today, of course, this is a problem of another dimension. In the metrical approach to phonology (see, for example, Giegerich 1985 and Goldsmith 1990), such rules are defined in terms of strong syllables and weak syllables. In Welsh, however, strength and weakness can follow two different parameters. In both of the syllable deletion rules, the syllable deleted precedes the accented or strong syllable; but in one (following tragywyddol) the syllable that is strong in amplitude triggers the deletion of the preceding weak syllable, while in the other (following eto) the metrically strong syllable in pitch is weaker in amplitude than the syllable deleted - the very syllable that by its strength motivates deletion as well.
It is probably this problem that causes B.J. Williams (1989) to define stress in Welsh in terms of pitch for the traditional phonological structure, while she avoids the role of amplitude in the penult for the more innovative elements of the structure. What we must do now is to find some way of reconciling the two in the single structure in which they now coexist.
Davies, Aled Lloyd. 1983. Cerdd Dant: Llawlyfr Gosod. Caernarfon: Gwasg Gwynedd. (Instrumental Music for Penillion: Handbook of the Gosod).
Giegerich, Heinz J. 1985. Metrical Phonology and Phonological Structure. Cambridge: Cambridge University Press.
Goldsmith, John A. 1990. Autosegmental and Metrical Phonology. Oxford: Blackwell.
Griffen, Toby D. 1977. "Deep and Surface Syllables". 52nd Annual Meeting, Linguistic Society of America, December 1977, Chicago
Griffen, Toby D. 1979. "On Phonological Stress in Welsh". Bulletin of the Board of Celtic Studies 28.106-112.
Griffen, Toby D. 1981. "Prosodic Alliteration in Cynghanedd Poetry". Bulletin of the Board of Celtic Studies 29.497-503.
Griffen, Toby D. 1991/92. "Epenthesis and the Old Welsh Accent Shift". Studia Celtica 26/27.163-74.
Gwynn Williams, W.S. 1971. Welsh National Music and Dance. 4th ed. Llangollen: Gwynn.
Jones, D.M. 1949. "The Accent in Modern Welsh". Bulletin of the Board of Celtic Studies 13.63-64.
Jackson, Kenneth H. 1975/76. "The Date of the Old Welsh Accent Shift". Studia Celtica 10/11.40-53.
Kinney, Phyllis. 1979. "The Tunes of 'Yr Hen Benillion'". Canu Gwerin/Folk Song 2.30-45.
Lloyd, D. Idwal. 1985. Celtic Word Craft. Trewolsta (Cornwall): Dyllansow Truran.
Morris Jones, John. 1913. A Welsh Grammar: Historical and Comparative, Phonology and Accidence. Oxford: Clarendon.
Morris Jones, John. 1925. Cerdd Dafod. Oxford: Clarendon. (The Art of Poetry)
Parry, Thomas. 1962. The Oxford Book of Welsh Verse. Oxford: Clarendon.
Parry Williams, D.E. 1948. "Music and Religion". Music in Wales ed. by Peter Crossley-Holland, 51-60. London: Hinrichsen.
Parry-Williams, T.H. 1936. Elfennau Barddoniaeth. Cardiff: University of Wales Press. (Elements of Poetry)
Polin, Claire. 1982. The ap Huw Manuscript. Henryville, PA: Institute of Mediaeval Music.
R. Roberts. 1914. "Penhillion Singing with the Harp, or, Canu Gydâ'r Tannau". Transactions of the Honourable Society of Cymmrodorion 148-65.
Roberts, R.E. 1925/26. "Welsh Music in the Tudor Period". Transactions of the Honourable Society of Cymmrodorion 1-24.
Rowlands, Eurys I. 1976. Poems of the Cywyddwyr: A Selection of Cywyddau c. 1375-1525. Dublin: Dublin Institute for Advanced Studies.
Saer, D. Roy. 1977. Carolau Plygain: Plygain Carols. Traddiodiad Gwerin Cymru: Welsh Folk Heritage 1. Fagans, Cardiff: Welsh Folk Museum.
Saer, D. Roy. 1980. Caneuon Llofft Stabal: Stable-Loft Songs. Traddiodiad Gwerin Cymru: Welsh Folk Heritage 2. Fagans, Cardiff: Welsh Folk Museum.
Strohm, Reinhard. 1993. The Rise of European Music, 1380-1500. Cambridge: Cambridge University Press.
Watkins, T. Arwyn. 1961. Ieithyddiaeth: Agweddau ar Astudio Iaith. Cardiff: University of Wales Press. (Linguistics: Aspects of the Study of Language)
Watkins, T. Arwyn. 1972. "The Accent in Old Welsh - its Quality and Development". Bulletin of the Board of Celtic Studies 25.1-11.
Williams, Briony J. 1989. Stress in Modern Welsh. Bloomington: Indiana University Linguistics Club.
Williams, J. Lloyd. 1909. Welsh National Melodies and Folk-Song. London: Honourable Society of Cymmrodorion.
Copyright © Toby D. Griffen, 2000
This edition copyright © Celtic Cultural Studies, 2000
ISSN 1468-6074
The moral, intellectual, and other universally-recognised copy rights
of the author are hereby registered and asserted under the terms of
UK, European Union, and other internationally valid copyright laws.
All rights reserved.
