2.4 The syllable and phonotactics
2.4.2 Consonant clusters
2.4.2.1 Tautosyllabic clusters
Tautosyllabic clusters are restricted to the onset of a syllable, no more than two consonants may occur and they only involve a subset of the phonemes. In a σ[C1C2 template, C2 may only be /r/ or /w/.
In a cluster with /r/ we find all consonant phonemes except for the three nasal stops (∗σ[mr, ∗σ[nr, ∗σ[Nr) and the approximants (∗σ[wr and ∗σ[yr) and /r/ itself (∗σ[rr). This points to an explanation in terms of a sonority hierarchy in which nasal and approximants are more sonorous than the trill/tap. Some ex- amples ofσ[Cr clusters arebrüzi ‘catfish type’, frar ‘small fishtrap’,krüfr ‘cold’,
gru ‘shooting star’, kwras ‘Brolga’, srima kabe ‘scout, spy’, thru thru ‘bamboo type’, trisi ‘scratch (v)’, zra ‘swamp’.
In a cluster with /w/ the restrictions on C1 are more severe and roots in which
it is attested are rare. We only find the following phonemes in C1 position: /k/,
/g/, /z/, /nz/, /th/, and /s/. The first two phonemes in the list pose a problem because one has find a distinction between a Cw cluster and the labialised velar stops /kw/ and /gw/. This is impossible to do for lexemes, but we find some evidence in a morphophonemic rule in §2.5.3 where the vowel /u/ is realised as [w] and becomes part of a σ[Cw cluster. Some examples of lexemes with σ[Cw onset clusters are: swäyé ‘anchoring place’, zwäf ‘luke-warm’, bzwär [mb˘@zwæR]
‘place name’.
2.4.2.2 Heterosyllabic clusters
Heterosyllabic clusters are much harder to pin down because - as we will see in §2.4.3 below - there are syllabicity alternations where a coda consonant may be- come an onset by inserting epenthetic schwa after which it breaks up the cluster. I will label the two consonants involved Ca (the coda of the first syllable) and Cb
(the onset of the following syllable).
We find that where Caand Cbare identical the consonants are never broken up
but always realised as geminates. The attested geminate patterns are described as a phonological rule in §2.3.1. These patterns exclude a number of logically possible geminates: labialized velar stops (/kw/ and /gw/), velar nasal (/N/), and all the prenasalised phonemes (/b/, /d/, /g/, and /nz/).7 Other heterosyllabic
7The labialized velar stop and the velar nasal may not occur as C
abecause these never occur
clusters are rather unrestricted. Table 2.7 presents the possible cluster types in Komnzo and Table 2.8 lists examples of these types.
Table 2.7 Heterosyllabic consonant clusters
oral labio-
/r/ stop prenasal8 nasal affricate fricative approx velar
/r/ ✓ ✓ n/a ✓ ✓ ✓ ✓ ✓
oral stop n/a ✓ n/a ✓ n/a ✓ ✓ ✓ prenasal n/a ✓ n/a ✓ n/a ✓ ✓ n/a
nasal ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
affricate n/a ✓ n/a ✓ n/a ✓ ✓ n/a fricative n/a ✓ n/a ✓ ✓ ✓ ✓ ✓ approx n/a ✓ n/a ✓ ✓ ✓ n/a n/a lab-velar n/a n/a n/a n/a n/a n/a n/a n/a
Table 2.8 Examples of attested heterosyllabic consonant clusters Ca Cb underlying phonetic english
representation realization
/r/ [+nasal] /ker.ma/ [keRma] ‘from the tail’ /tr.nä/ [t˘@Rnæ] ‘palm frond’ /r/ [+oral stop] /for.tu/ [F˘ÅRtu] ‘scar’
/ker.ko/ [keRko] ‘type of headdress’ /r/ [+affricate] /zr.zü/ [ts˘@RtSY] ‘knee’
/r/ [+fricative] /war.fo/ [waRFO] ‘above’ /kr.si/ [k˘@Rsi] ‘block (v)’ /tr.tha/ [t˘@RDa] ‘life’ /r/ [+approx] /kar.wä.si/ [kaRwæsi] ‘lie, trick’
/yar.yom.g.si/ [jaRj˘ÅmNg˘@si] ‘scream (v)’
/r/ [+lab-velar] /Na.far.kw.re/ [NaFaRkw˘@Re] ‘we leave’
[+oral stop] [+oral stop] /wät.ku/ [wætku] ‘Australian Pelican’ [+oral stop] [+nasal] /dek.ni.ni/ [ndeknini] ‘praying mantis’
/rt.maksi/ [R˘@tmak˘@si] ‘cut’ [+oral stop] [+fricative] /f.rk.thé/ [F˘@R˘@kD˘@] ‘red’
/et.fth/ [PetF˘@T] ‘sleep (n)’
8The column and line labelled ‘prenasal’ includes prenasalised stops and the prenasalised
Ca Cb underlying phonetic english representation realization
[+oral stop] [+approx] /thik.ya.si/ [Dikjasi] ‘tie the fence’ /zok.wa.si/ [ts˘Åkwasi] ‘speech’ /mit.wa.si/ [mitwasi] ‘swing (v)’ [+oral stop] [+lab-velar] /tat.kwo.nam/ [tatkwOnam] ‘tree type’ [+prenasal] [+oral stop] /gb.ka.rä/ [Ng˘@mbkaRæ] ‘with pandanus’ [+prenasal] [+nasal] /Nad.me/ [Nantme] ‘with the rope’
[+prenasal] [+fricative] /bad.fo/ [mbantFO] ‘to the ground’
[+prenasal] [+approx] /mnz.wä/ [m˘@ntswæ] ‘really the house’
[+nasal] /r/ /nin.rr/ [ninR˘@R] ‘with us’ [+nasal] [+oral stop] /am.kf/ [Pamk˘@F] ‘breath’
/thun.t.nä.gw.rm/ [Dunt˘@næNgw˘@R˘@m] ‘he was losing
them here’ [+nasal] [+nasal] /kan.motha/ [kanmODa] ‘river snake’ [+nasal] [+prenasal] /yar.yom.g.si/ [jaRj˘ÅmNg˘@si] ‘scream (v)’
/kum.da/ [kumnda] ‘basket’
/kän.brim/ [kænmbRim] ‘come back here!’
[+nasal] [+affricate] /san.zin/ [santSin] ‘put him down here!’ [+nasal] [+fricative] /zan.fr/ [tsanF˘@R] ‘far’
/kam.tha.tha/ [kamDaDa] ‘like a bone’ [+nasal] [+approx] /nze.nm.wä/ [ndZen˘@mwæ] ‘really for us’
[+nasal] [+lab-velar] /Nan.kwir/ [Nankwir] ‘he run hither’
[+affricate] [+oral stop] /ez.kn.wr/ [Petsk˘@nw˘@R] ‘s/he moves them’ [+affricate] [+nasal] /käz.nob/ [kætsn˘Åmp] ‘drink (it)!’
[+affricate] [+fricative] /fz.fo/ [F˘@tsFO] ‘to the forest’ [+affricate] [+approx] /fz.wä/ [F˘@tswæ] ‘really the forest’ [+fricative] [+oral stop] /mnz.wä/ [m˘@ntswæ] ‘really the house’
[+fricative] [+affricate] /buf.zenz/ [mbuFtSents] ‘your wife’
[+fricative] [+fricative] /ef.thar/ [PeFDaR] ‘dry season’ /füs.füs/ [FYsFYs] ‘wind’
Ca Cb underlying phonetic english representation realization
[+fricative] [+approx] /nzf.wi.ya.km/ [nts˘@FwIjak˘@m] ‘we were walking’
/naf.wä/ [naFwæ] ‘really they (erg)’
/fith.wo.g.si/ [FiTwONg˘@si] ‘take out’
[+fricative] [+lab-velar] /math.kwi/ [maTkwi] ‘personal name’
/y.ra.kth.kwa/ [j˘@rak˘@Tkwa] ‘s/he put him on top’
[+approx] [+oral stop] /faw.ka.rä/ [FawkaRæ] ‘with a payment’
[+approx] [+nasal] /faw.ma/ [Fawma] ‘from the payment’
[+approx] [+affricate] /bäw.zö/ [mbæwtSœ] ‘paperbark’
[+approx] [+fricative] /wy.thk/ [w˘@jD˘@k] ‘it comes to an end’
We can make a number of observations from Table 2.8 above. The prenasalised phonemes do occur in Ca as well as Cb. In the latter case, Ca may only be an-
other nasal as in: kumda [kumnda] ‘basket’, kumgsi [kumNg˘@si] ‘smell (v)’, dmgu
[nd˘@mNgu] ‘waterhole’, tingwä [tinNgwæ] ‘tree type’. If Ca is a phoneme other
than a nasal, the cluster will be broken up: garda [NgaR˘@nda] ‘canoe’, äthgam
[PæD˘@Ngam] ‘Parinari nonda’, thfgarwrmth [D˘@F˘@NgaRw˘@R˘@m˘@T] ‘they were break-
ing them’. There are no attested cases of a prenasalised phoneme in Cb with a
homorganic nasal in Ca, i.e. /m/+/b/, /n/+/nz/, /n/+/d/.
There are only few clusters which involve /r/ in the Cb position. This is
caused by maximizing onsets during syllabification, which creates complex on- sets clusters of the type CR. As a consequence, the only heterosyllabic clusters with /r/ in Cbposition are the ones which are illegal as onset clusters (e.g. ∗σ[mr,
∗
σ[nr, ∗σ[rr). In other words, because ∗σ[nr is illegal as an onset, we do find it as a heterosyllabic cluster (ninrr /nin.rr/ [ninR˘@R] ‘with us’). Likewise, becauseσ[fr is a legal onset cluster, we never find it as a heterosyllabic cluster.
We do find heterosyllabic clusters which involve /w/ in Cb position and a
velar (prensalized) stop in Ca position. Evidence that these clusters are indeed
heterosyllabic as opposed to an instantiation of the labialized velar stop /kw/ and /gw/ comes two sources. First, we find examples like zokwasi [ts˘Åkwasi] ‘speech’ where the short, centralized allophone of /o/ shows that /k/ is the coda of a closed syllable. (Compare with the discussion of /o/ in §2.2.1 and the discussion of syllable weight in §2.4.1 above) Second, in verb stems ending in /k/ and /g/ the following non-dual suffix is always -wr (§5.5.3.3). Both cases attest that we find heterosyllabic /k.w/ and /g.w/ clusters in addition to the complex phonemes /kw/ and /gw/.
2.4.3
Syllabification and epenthesis
Syllable structure is generally understood not to be defined at the underlying rep- resentation (Blevins, 1995: 221). Hence, we do not find minimal pairs based on syllabicity in Komnzo. As was explained in §2.2.2 above, schwa is not a phoneme but an epenthetic vowel inserted in order to break up consonant clusters. There is some degree of free variation in syllabicity and schwa insertion. An example is the word mrn ‘family, clan’ with the locative suffix -en. The resulting word
mrnen ‘in the family’ may be realised either /mr.nen/ [m˘@Rnen] or /m.r.nen/ [m˘@R˘@nen]. There is no phonemic contrast and speakers find it difficult to per- ceive the difference in syllabicity.
The process of syllabification will be outlined here in the form of three ordered rules which predict epenthesis and syllable structure:
1. Associate each specified vowel with a syllable nucleus.
2. Establish and maximize onsets in accordance with syllable templates (See constraint number 2 in §2.4.1 on onset clusters). A phonological rule will insert a glottal stop if there is no consonantal onset in word initial position (See §2.3.3).
3. Break-up unsyllabified consonants with epenthetic vowels:
(a) Exception: suffixes which allow no other syllabification than inserting the epenthetic vowel in final position. This includes the adjectivalizer
-thé, non-singular ergative case marker-yé and the first singular actor verb suffix-é.
(b) Elsewhere: proceed from right to left breaking up consonant clusters. (c) After each schwa insertion, establish codas in accordance with pos- sible heterosyllabic consonant clusters. Otherwise, maximize onsets. Exception: word-initial segments are always recognized as onsets. (d) The epenthetic vowel is [˘u] and [˘I] if followed by heterosyllabic /w/
and /y/ respectively. In all other instances it is [˘@].
The process of syllabification attempts to map the minimal syllable CV onto the underlying representation. The rules give preference to onsets rather than codas. Consequently, we do not find vowel initial syllables word-medially or word- finally.
I have modelled the process of syllabification as being divided into two steps. Syllables which contain full vowels are recognized first and in a second step epenthetic vowels are inserted to break up unsyllabified consonant clusters. This algorithm proceeds backwards (from right to left) and inserts epenthetic schwas
between unsyllabified consonants to create syllable nuclei. The insertion ensures that onsets are maximized. After each onset, the processs checks against the list of possible heterosyllabic consonant clusters (§2.4.2.2) whether another insertion occurs right away or only after a coda has been recognized. In the latter case, it ‘jumps’ one consonant and breaks up the next pair of unsyllabified consonants. An exception is the word initial position where the segment is automatically rec- ognized as an onset. The rules ensure that no word-initial schwa insertion occurs. The direction (right to left) explains why we find schwa never in word-final po- sition. There are only a handful of lexemes in which schwa is attested word-finally.
The direction is important in order to explain forms likewonrsoknwr [w˘Ån˘@RsO- k˘@nw˘@R]9 ‘s/he is bothering me’ which is syllabified /wo.nr.so.kn.wr/. The algo- rithm is applied from right to left which is why the cluster /r.s/ is first recognized as a possible heterosyllabic consonant cluster. After this recognition, schwa is in- serted between /n/ and /r/. If the process was applied from left to right, one would expect that /n.r/ is first recognized as a possible heterosyllabic cluster and schwa would be inserted between /r/ and /s/ which yields the incorrect form
∗/won.r.so.kn.wr/. As pointed out above, there is some degree of optionality. In
elicitation, informants accepted schwa insertion in both places [w˘Ån˘@R˘@sOk˘@nw˘@R]. This might be an artefact introduced by elicitation, because in fluent speech this hardly ever occurs.
The algorithm specifies that schwa is inserted between consonants disregard- ing possible onset clusters (§2.4.1) whereas syllables with specified vowels maxi- mize their onsets and produce onset clusters. Indeed, we do not find the possible onset clusters CR or CW with epenthetic vowels. There are only two exceptions for CR. The first is the verb frmnzsi /frm.nz.si/ ‘fix, prepare’ in which the on- set cluster /fr/ is never broken up even if the verb is fully inflected: yafrmnzr
/ya.frm.nzr/ ‘s/he prepares him’. The second exception occurs with all verbs in a specific inflection: Word-initially, the irrealis prefixr-becomes part of an onset cluster with the undergoer prefix. This cluster only contains an epenthetic vowel if (i) the restricted verb stem is used and (ii) the verb is marked for dual number:
thrthbth [DR˘@D˘@mb˘@T] ‘they put them inside’.10
In Figure 2.3-2.5 below, I present four examples spelling out the algorithm step by step:
9The allophone [˘Å] of the phoneme /o/ occurs here not because this might be a closed
syllable, but because it follows a labio-velar approximant (See §2.2.1)
10This verb is glossed as: th-r-∅-thb-th 2|3nsg-irr-nd-put.inside.rs-2|3nsg It it a rare
inflection because three things have to come together: irrealis mood, restricted verb stem, dual number marker (which is a zero-morpheme in this case).
Figure 2.3 Syllabification ofkwark ‘deceased’
/kwark/ underlying representation ↓
/kwσ[a]rk/ Rule 1: Associate each specified vowel with a nucleus.
↓
/σ[kwa]rk/ Rule 2: Maximize onsets. →establishes the syllableσ[kwa]
↓ /σ[kwa]σ[rk]/
Rule 3b: Break up consonant clusters. →schwa is inserted bet- ween /r/ and /k/ and creates a CVC syllable
↓
/kwa.rk/ syllabified form ↓
[kwa.R˘@k]
Figure 2.4 Syllabification ofyanthugwr ‘s/he tricks him here’
/yanthugwr/ underlying representation ↓
/yσ[a]nthσ[u]gwr/ Rule 1: Associate each specified vowel with a nucleus.
↓
/σ[ya]nσ[thu]gwr/ Rule 2: Maximize onsets. → establishes the syllables .ya.
and .thu. ↓ /σ[ya]nσ[thu]gσ[wr]/
Rule 3b: Break up consonant clusters. →schwa is inserted between /w/ and /r/ because it is not a possible cluster
↓ /σ[ya]nσ[thug]σ[wr]/
Rule 3c: Establish codas in accordance with possible heterosyllabic consonant clusters. → /g.w/ is possible → /g/ becomes a coda of the preceding syllable.
↓ /σ[yan]σ[thug]σ[wr]/
Rule 3c: Establish codas in accordance with possible heterosyllabic consonant clusters. →/n.th/ is possible → /n/ becomes a coda of the preceding syllable.
↓
/yan.thug.wr/ syllabified form ↓
Figure 2.5 Syllabification ofzwäfiyokwé ‘I finished sth. for her’
/zwäfiyokwσ[é]/
underlying representation: the final schwa is the1sgac-
tor suffix. It is must be prespecified as syllabic in the underlying representation
↓
/zwσ[ä]fσ[i]yσ[o]kwσ[é]/ Rule 1: Associate each specified vowel with a nucleus.
↓
/σ[zwä]σ[fi]σ[yo]kσ[wé]/ Rule 2: Maximize onsets. → establishes the syllables
[thwä], [fi], [yo], and [wé] ↓
/σ[zwä]σ[fi]σ[yok]σ[wé]/
Rule 3c: Establish codas in accordance with possible heterosyllabic consonant clusters. → /k.w/ is possible →/k/ becomes the coda of the preceding syllable.
↓
/zwä.fi.yok.wé/ syllabified form ↓
[tswæ.BI.j˘Åk.w˘@]
Figure 2.6 Syllabification ofskrifzenz ‘Skri’s wife’
/skrifzenz/ underlying representation ↓
/skrσ[i]fzσ[e]nz/ Rule 1: Associate each specified vowel with a nucleus.
↓
/sσ[kri]fσ[ze]nz/ Rule 2: Maximize onsets. →establishes the syllables .kri. and
.ze. ↓ /sσ[kri]fσ[zenz]/
Rule 3c: Establish codas in accordance with possible hetero- syllabic consonant clusters. → no cluster with /nz/ → /nz/ becomes the coda of the preceding syllable.
↓ /sσ[krif]σ[zenz]/
Rule 3c: Establish codas in accordance with possible hetero- syllabic consonant clusters. →/f.z/ is possible→/f/ becomes the coda of the preceding syllable.
↓ /σ[s]σ[krif]σ[zenz]/
Rule 3b: Break up consonant clusters. → schwa is inserted between /s/ and /k/ because it is not a possible cluster
↓
/s.krif.zenz/ syllabified form ↓