• Dutch
  • Frisian
  • Afrikaans
Show full table of contents
The syllabic affiliation of prevocalic glides

This section deals with the syllabic affiliation of glides preceding a vowel (short, long, diphthong), viz. with the question as to whether these glides belong to the syllable onset or to the syllable nucleus.


In the late 17th and early 18th centuries, the long central vowels e:,o:,ɛ:,ɔ: were diphthongized to iɪ,ɪɛ,uo,oɔ. These diphthongs then began to be realized as either [iə,ɪə,uə,oə] or [jɪ,jɛ,wo,wa], depending on the morphophonological context. This sound change is known as Modern Frisian Breaking (see Breaking and the references cited there). It resulted in a huge increase in initial consonants and consonant sequences followed by the glides [j] and [w], which no doubt is a striking feature of Frisian phonotactics.

The phonological interpretation of these sequences is a matter of debate. The question is whether the glides they contain belong to the syllable onset, the syllable nucleus, or to neither of the two. The question behind this is whether or not Frisian has rising diphthongs. In Cohen et al. (1959:125-126), the glide in the above clusters is considered to be a (consonantal) onset phoneme, since 1) it can be replaced by another consonant, rendering minimal pairs, like jûk[juk]yoke - tûk[tuk]skilful and clever and jier[jiər]year - dier[diər]animal, and 2) it does not constitute a syllable. This means that no rising diphthongs are assumed to exist. As a consequence, triconsonantal initial clusters become quite numerous. This view is adopted by Booij (1989).

At first sight, this is a desirable outcome in view of what is stated in Kenstowicz and Rubach (1987:476). They observe that the glide of a diphthong "is characteristically oriented in the same direction with respect to the core throughout the entire system of diphthongs, as either onglide (Slovak) or offglide (English, Canadian, French). It is natural to construe this left-right orientation as fixing a parameter of the representational system". Since Frisian has genuine left-headed (falling and centring) diphthongs, right-headed (rising) ones are not expected to occur, so the glides in the above-mentioned clusters should preferably not be interpreted as part of the nucleus.

Segments within one and the same phonological constituent can be subjected to collocational restrictions, as stated in Ewen and Van der Hulst (2001:130):

"Further evidence for the validity of the onset-rhyme division has been found in the apparent independence of the two constituents. That is, on the assumption that a syllable can be seen as a sequence of onset and rhyme, it has been claimed that the constraints on the co-occurrence of segments holding between onset and rhyme are much less severe than those holding within each of the two constituents. That is, given a list of well-formed onsets and well-formed rhymes, these can combine quite freely to form well-formed syllables. Thus onsets and rhymes are seen as autonomous units, each with their own constraints on their internal structure."

Collocational restrictions can thus shed light on what constituent a given segment belongs to. In Dutch, for instance, the glides [j] and [w] "precede all kinds of vowels, and are subject to co-occurence restrictions with preceding consonants" (Booij (1989:321)), cf. Cohen et al. (1959:88-89)). This means that the Dutch glides are best analyzed as belonging to the onset, in line with the fact that Dutch has falling diphthongs.

In Frisian, however, matters are not that straightforward. On the one hand, the glides [j] and [w] co-occur with all sorts of tautomorphemic vowels (short ones, long ones, diphthongs) to their right. On the other hand, they combine with virtually all permissible tautomorphemic consonants and consonant clusters to their left. An overview of the possibilities is presented in tabel (A) and (B):

Table 1: Tabel (A): single consonants
poask[pwask]ruff puol[pwol]pool
pjisk[pjɪsk]peach pjuts[pjøts]gush, spurt
boarst[bwast]breast buol[bwol]bubble
bjist[bjɪst]beestings bjuster[bjøstr̩]lost
twer[twɛr]sickening twir(je)[twɪr]to whirl
tjirk[tjɪrk]redshank tjems[tjɛms]milk filter
dwers[dwɛs]crosswise dwyl[dwil]dizzy
djip[djɪp]deep djerre[djɛ:rə]yolk
kwelts[kwɛlts]with a limp kuos[kwos]pet name of pig
kiuw[kjo:]gill kjirm(je)[kjɪrm]to moan
Goasse[ɡwasə]proper noun guod[ɡwot]things
gjin[ɡjɪn]no(ne) gjalp[ɡjɔlp]shriek
fuort[fwot]away foars[fwas]sturdy
fjoer[fju.ər]fire fjild[fjɪlt]field
woarst[vwast]sausage woartel[vwatl̩]root
wiette[vjɪtə]wetness wjok[vjok]wing
swilk[swɪlk]linoleum swel[swɛl]swallow
sjerp[sjɛrp]syrup sjippe[sjɪpə]soap
loarte[lwatə]droppings luork(je)[lwork]to peek
ljisk[ljɪsk]groin ljurk[ljørk]lark
ruot[rwot]soot roaster[rwastr̩]grid
rjocht[rjoxt]law, justice rjemme[rjɛmə]cream
noartsk[nwatsk]surly nuodlik[nwodlək]risky
njonken[njoŋkŋ]next to njúnt(sje)[njynt]to hum
muoz(je)[mwoz]to spill muoi(e)[mwo:j]to regret
meall(e)[mjɛl]to grind mjitte[mjɪtə]measure

Table 2: Table (B1): clusters of obstruent + liquid
pljims[pljɪms]sheet (of paper); quarter of a quire
([plw] is not attested)
priuw(e)[prj{o:/u:/y:}wə]to taste
preamk(je)[prjɛmkjə]sail with a flatboat
proalling[prwalɪŋ]kidney of a slaughtered animal
proastich[prwastəx]jaunty, perky
bliuw(e)[blj{o:/u:/y:}wə]to stay
blier(je)[bljɪrjə]to blister, to bubble
bluodderich[blwodərəx]bloody, gory
brjit[brjɪt]lump of peat
briefk(je)[brjɪfkjə]to correspond (with)
bruorren[brworn̩]brothers; brethren
triuw(e)[trj{o:/u:/y:}wə]to push
triemmich[trjɪməx]rancid, rank
troanje[trwãjə]face; mug, phiz
driuw(e)[drj{o:/u:/y:}wə]to float, to drift
([drw] is not attested)
kljirre[kljɪrə]tangle (of hair)
kluork(je)[klworkjə]to question, to sound out
kloark(je)[klwarkjə]to question, to sound out
kreauw(e)[krjo:wə]to quarrel
kriuwel[krjo:wəl]itch, tickle
kroask(je)[krwaskjə]to pace reluctantly
glierk(je)[ɡljɪrkjə]to shine, to glance
gljurk(je)[ɡljørkjə]cast loving looks at
gluork(je)[ɡlworkjə]to gleam, to glow
gloarje[ɡlwarjə]glowing ember
groatten#brij[ɡrwatn̩-]barley gruel
frjemd[frjɛmt]foreign; strange
Frjentsjer[frjɛntsjər]name of a Frisian town
fljoch[ljox]fly (present tense stem)
fljirk(je)[fljɪrkjə]to flame, to kindle
fluor(je)[flworjə]to pave, to surface
wriuw(e)[vrjo:wə]to rub
([vrw] is not attested)

Table 3: Table (B2a): biconsonantal clusters
fnjit (hawwe mei)[fnjɪt](be) pleased (with)
gnjird(zje)[ɡnjɪdzjə]seize roughly
gnoar(je)[ɡnwarjə]to growl; to grumble
knjocht[knjoxt]servant, helper
in knoarre[knwarə]a lot
smjunt[smjønt]wigeon; rascal, scoundrel
smjirk(je)[smjɪrkjə]make a mess
smoarch[smwarx]dirty, soiled
(op it) snjit (mei)[snjɪt]pleased (with)
snuorje[snworjə]certain period of time
snoark(je)[snwarkjə]to snore
sloarm[slwarm]great lump of a dog
sluork(je)[slworkjə]to slide softly
sljurk(je)[sljørkjə]to slide softly
sljocht[sljoxt]even, smooth
spoanne[spwanə]chip (of wood)
spuonnen[spwonn̩]made of split-wood
stoarm[stwarm]storm, gale
stuolk(je)[stwolkjə]be based (on)
stjitt(e)[stjɪtə]to thrust
stjelp[stjɛlp]cheese-cover farmhouse

Table 4: Tabel B2b: triconsonantal clusters (initiating in /s/)
strjemm(e)[strjɛmə]to coagulate, to curdle
spriuw[sprjo:]thrush, aphtha
skriuw(e)[skrjo:wə]to write
skreau[skrjo:]shout, shriek
skroar(je)[skrwarjə]practise the tailor's trade

In view of all this it seems that we have to face a contradiction. There appears to be evidence that glides do not belong to the nucleus, viz. their co-occurrence with all sorts of vowels. At the same time, however, there appears to be evidence that they do not belong to the onset either, viz. their co-occurrence with all sorts of consonants and consonant clusters. Put differently, glides do not seem to contribute to the complexity of either nucleus or onset.

[show extra information]

Hermans (2007) notes that [ju] and [jy] can only be preceded by a consonant cluster when followed by the glide [w], as in bliuw(e)/blj{o:/u:/y:}/to stay and triuw(e)/trj{o:/u:/y:}/to push. In the inflected forms bliuwe[blj{o:/u:/y:}wə]stay (infinitive; all plural persons present tense) and triuwe[trj{o:/u:/y:}wə]push (infinitive; all plural persons present tense), however, the glide [w] functions as a hiatus filler (see the resolution of hiatus between a monophthong and a following vowel), so it is not part of the underlying representation of the verbs at hand. The spelling, however, seems to suggest that [w] is part of the stem, since words like these are always written with <iuw>.

[show extra information]

On closer scrutiny, there appears to be a small asymmetry between onset + [j/w] and [j/w] + nucleus. On the one hand, the glides are not followed by all vowels and vowel combinations: [j] does not precede [ɪə, øə, o:j, ɔ:] and [w] does not precede [i:w, y, y:, yə, ø:, öə, o:, oə, ʌy]. On the other hand, they cannot be preceded by any consonant combination either: [w] cannot combine with the initial clusters [dr, pl, vr, fn, skl, spr, str, spl], whereas [j] cannot with [skl, spl]. The glides do not behave symmetrically here, since the combinatorial possibilities of [j] exceed those of [w].

The three words in (1) are an exception to the above statement that [j] does not precede [ɪə]:

Example 1

tsjea [tsjɪə] thigh
tsjeaf [tsjɪəf] thief
tsjeak [tsjɪək] jaw; cheek

These words, however, have become obsolete. As to this, it should be added that the sequences [jøə, jy, jɪə] did occur as well, as the words in (2) show:

Example 2

frjeon [frjøən] friend [(nowadays: freon)]
krjús [krjys] cross [(nowadays: krús)]
ljeaf [ljɪəf] dear, nice [(nowadays: leaf)]

The sequences [jyə] and [jɪə] show dialectal variation in [yə] and [ɪə]:

Example 3

njuet ~ nuet [n(j)yət] tame(d)
sjeas ~ seas [s(j)ɪəs] gig
[show extra information]

Hermans (2007) analyzes the products of Modern Frisian Breaking as monopositional vowels with two place nodes, hence as complex vowels. In his analysis then the glides are part of the nucleus. This is also the position taken by Visser (1997:Chapter 3) and Visser (2002). According to the analysis put forward in Booij (1989) and Chada (2007), on the other hand, they are part of the onset, which entails that Frisian only has falling and centring diphthongs, but not rising ones.

It seems obvious then that more conclusive evidence will have to be produced on the basis of which one can make plausible whether a glide belongs to the nucleus or to the onset. There appears to be evidence in favour of either analysis. The evidence which points to a glide belonging to the onset will be presented first.

In the first place, a language game seems to provide evidence for the onset analysis. The game at hand entails the hypocoristic transformation of personal names, exemplified in the scheme below:

Table 5
Krijn[krɛjn] Krijn petijn[krɛjm pətɛjn]
Douwe[dɔwə] Douwe petouwe[dɔwə pətɔwə]
Richt[rɪxt] Richt peticht[rɪxt pətɪxt]
It is clear what this game consists of: the name is repeated, in which the onset is replaced with the sequence [pət-], a kind of reduplication. Now, names beginning with a glide or the sequence consonant + glide, like Jarich[ja:rəx], Lieuwe[ljo:wə], Sjoerd[sjuət], Goasse[ɡwasə], and Duotsje[dwotsjə] are transformed as shown in the table below:

Table 6: Examples of the transformation of names beginning with a glide or the sequence consonant + glide
With the glide [j] With the glide [w]
Jarich petarich[pəta:rəx][*pətja:rəx] Goasse petoasse[?pətasə][??pətwasə]
Lieuwe petieuwe[pəto:wə][*pətjo:wə] Duotsje petuotsje[?pətotsjə][??pətwotsjə]
Djoke petoke[pəto:kə][*pətjo:kə]
Sjoerd petoerd[pətu.ət][*pətju.ət]

The pattern in the left-hand column seems to indicate that the glide belongs to the onset. The intuitions about the cases in the right-hand column are not that clear. Though there may be a slight preference for the forms without a glide, those with a glide are not ill-formed.

[show extra information]

As to the pattern in the left-hand column, it should be noted that the sequence [tj] is rather uncommon in Frisian. It occurs mainly in complex words with a broken centring diphthong, like teannen[tjɛnn̩]toes (cf. teantɪəntoe) and tiennen[tjɪnn̩]wicker (cf. tien/tiən/osier (rod), wicker). As for simplex words, it only occurs in the unfamiliar verb tjild(e)/tjɪld/to recognize, to acknowledge and some interjections, like tjú(jút)[tjy(jyt)], tjút[tjyt], and tjûk[tjuk]. All other words beginning with [tj-] have a variant with [tsj-], so tjilling[tjɪlɪŋ]teal and tjirk[tjɪrk]redshank have the variants tsjilling and tsjirk. The other way around, however, not all words beginning with /tsj/- have a variant with [tj-], so the variant forms *tjaffel and *tjerke do not occur alongside tsjaffel[tsjafəl]trap, gob and tsjerke[tsjɛrkə]church. Moreover, the sequence [*tjə] is out, only [tsjə] being allowed.

Secondly, in the eastern part of the language area, the sequences [wa] and [wo] have turned into [ja] and [jo] when preceded by a labial consonant (see replacement of the glide /w/ of the broken diphthong /w{a/o}/ by /j/ following labial consonants). Examples are provided in (4):

Example 1

Examples of words which show the change from [w{a/o}] to [j{a/o}] when following a labial consonant
boartsje [bjatsjə] to play [(elsewhere: [bwatsjə])]
foarke [fjarkə] fork [(elsewhere: [fwarkə])]
poarte [pjatə] gate [(elsewhere: [pwatə])]
woarst [vjast] sausage [(elsewhere: [vwast])]
muorre [mjorə] wall [(elsewhere: [mworə])]
fuort [fjot] gone; away [(elsewhere: [fwot])]
buorren [bjorn̩] village centre [(elsewhere: [bworn])]

The impetus for this change is a constraint prohibiting a sequence of two labial segments in the onset. This would imply that the glides must have been part of the onset.

In what follows, the evidence which points to a glide belonging to the nucleus will be presented.

In the first place, if the onsets of two successive words are interchanged, the glide remains in place, which is exemplified in the scheme below:

Table 7
Examples of the interchanging of onsets
wy hawwe om 'e keallen tocht[kjɛln̩ tɔxt]we have of the calves thought
as: wy hawwe om 'e teallen kocht[tjɛln̩ kɔxt]
*wy hawwe om 'e tellen kjocht[tɛln̩ kjɔxt]
wy hawwe de buorden makke[bwodn̩ makə]we have the signs made
as: wy hawwe de muorden bakke[mwodn̩ bakə]
*wy hawwe de morden boakke[modn̩ bwakə]
The fact that the glide remains in place seems to be an indication that it belongs to the nucleus.

Secondly, Modern Frisian Breaking (see breaking) resulted in an alternation between centring and rising diphthongs: /iə/ ~ /jɪ/, /uə/ ~ /wo/, /ɪə/ ~ /jɛ/. The diphthong /oə/, however, does not fit into this pattern, for it does not alternate with /wɔ/ as expected, but with /wa/. Hoekstra (1988) posits the phonotactic filter below, for which he adduces several pieces of evidence:

breaking filter

On the strength of this filter, the outcome of the breaking of /oə/ need to be repaired, which is achieved by means of lowering /ɔ/ to /a/. Since it is predominantly segments within one and the same phonological constituent that can be subjected to collocational restrictions, the very existence of the Breaking Filter indicates that /w/ and /ɔ/ are likely to belong to the same phonological constituent, viz. the nucleus.

Thirdly, the initial cluster /ts/ has the distributional property that it must precede the vowel /i/ (/i/, /i:/, /iə/), or the glide /j/, which is /i/ underlyingly (see the glides). This can be expressed by means of the following morpheme structure constraint (where 'M' is an abbreviation of 'Morpheme'):

if M(ts, then M(tsi

The distribution of word-initial /ts/ is exemplified in (5):

Example 2

Examples of words with the initial cluster /ts/
tsiis /tsi:z/ cheese
tsien /tsiən/ ten
tsjil /tsjɪl / wheel
tsjêf /tsjɛ:v/ chaff
tsjoar /tsjoər/ tether

V(V) and jV(V) pattern alike here, which is an indication that the glide belongs to the nucleus.

[show extra information]

Evidence which is neutral with respect to the nucleus- or onset-analysis ‒ see Visser (2002) ‒ is left out of consideration.

All evidence taken together, it turns out to be the case that co-occurrence restrictions are not always as unequivocal an indication concerning constituenthood as one might wish. In the case at hand, the prevocalic glides show ambiguous behaviour, in that there is evidence that they are part of the syllable onset and the syllable nucleus. Visser (2000) suggests therefore that these glides are floating segments in underlying representation: initially they remain unsyllabified, but at a later stage they are incorporated into the already existing prosodic structure. The latter may proceed by means of leftward or rightward association, viz. incorporation into the onset or into the nucleus, which would account for the ambiguous behaviour of these glides.

Frisian is not the only language in which the syllabic affiliation of prevocalic glides is problematical. If the conception of the syllable as the conjunction of an onset and a rhyme is taken seriously, then a) each segment must uniquely belong to either onset or rhyme and b) the boundary between these constituents should be clear and consistent for any given language, see Yip (2003), who considers prevocalic glides as a test-case for this set of assumptions. She notes that in English and Mandarin Chinese such glides behave as part of the onset and as part of the rhyme. As to this, there is both inter- and intra-speaker variation. As this casts doubt on the usefulness of the onset-rhyme distinction, a 'flat' syllable structure is opted for. The latter implies "simple, linear phonotactics, with no appeal to sub-syllabic constituency". Co-occurrence facts between segments are to be understood in terms of general notions like 'similarity', 'proximity', and 'sonorancy'. This view of the syllable seems to be corroborated by the behaviour of the prevocalic glides in Frisian.

  • Booij, Geert1989On the representation of diphthongs in FrisianJournal of Linguistics25319-332
  • Booij, Geert1989On the representation of diphthongs in FrisianJournal of Linguistics25319-332
  • Booij, Geert1989On the representation of diphthongs in FrisianJournal of Linguistics25319-332
  • Chada, Agnieszka2007Fake diphthongs in Frisian: an Optimality Theory approachAnglica16185-198
  • Cohen, Antonie, Ebeling, C.L., Eringa, P., Fokkema, K. & Holk, A.G.F. van1959Fonologie van het Nederlands en het Fries: Inleiding tot de moderne klankleerMartinus Nijhoff
  • Cohen, Antonie, Ebeling, C.L., Eringa, P., Fokkema, K. & Holk, A.G.F. van1959Fonologie van het Nederlands en het Fries: Inleiding tot de moderne klankleerMartinus Nijhoff
  • Ewen, Colin J. & Hulst, Harry van der2001The Phonological Structure of Words: An IntroductionCambridge University Press
  • Hermans, Ben2007Friese stijgende diftongen vormen een complex segmentPhilologia Frisica anno 2005 : lezingen en neipetearen fan it ... Fryske filologekongres1733-62
  • Hermans, Ben2007Friese stijgende diftongen vormen een complex segmentPhilologia Frisica anno 2005 : lezingen en neipetearen fan it ... Fryske filologekongres1733-62
  • Hoekstra, Jarich1988Wêrom brekt [o.ə] ta [wa]?Tydskrift foar Fryske Taalkunde448-53
  • Kenstowicz, Michael & Rubach, Jerzy1987The Phonology of Syllabic Nuclei in SlovakLanguage63463-497
  • Visser, Willem1997The Syllable in FrisianVrije Universiteit AmsterdamThesis
  • Visser, Willem2000Frjemd wurdt eigener. Oer de âlde Frânske lienwurden yn it FryskIt Beaken62141-218
  • Visser, Willem2002Brekking as ferkoarting: in diagroane analyzeIt beaken : meidielingen fan de Fryske Akademy Oer 'brekking'64187-246
  • Visser, Willem2002Brekking as ferkoarting: in diagroane analyzeIt beaken : meidielingen fan de Fryske Akademy Oer 'brekking'64187-246
  • Yip, Moira2003Casting doubt on the Onset~Rhyme distinctionLingua113779-816
Suggestions for further reading ▼
  • Dutch
  • Frisian
  • Afrikaans
Show more ▼
  • Dutch
  • Frisian
  • Afrikaans
Show more ▼
  • Dutch
  • Frisian
  • Afrikaans
Show more ▼