In complex segments and complex segments as single units, /ts/ has been treated on a par with the other /s/ + plosive (and plosive + /s/) clusters, which consist of two segments in underlying representation and surface as complex segments. A separate discussion of /ts/, however, is justified, for there are indications that this cluster might be viewed as an affricate (see Lin (2011) on affricates in general).

In the first place, /ts/ has often developed out of /k/, see the examples in (1):

On the basis of the above pattern, it might be concluded that /ts/ is a single segment. It should be kept in mind, though, that /ts/ in (1) is the outcome of a historical process of palatalization and assibilation, which is no longer productive in Frisian.

Secondly, the pairs of words in (2) — which have an unclear dialectal distribution — only differ in the presence of /ts/ vs. /k/:

In these words as well, /ts/ has developed out of /k/. The forms with /ts/ are becoming obsolete (see also Tamminga (1987)). In Standard Frisian, however, the form elts is more or less being propagated, because of distancing from Dutch, which has elk here. But on no account is there a productive pattern of alternation here.

Thirdly, hypocoristics may have /k/ where the full form of the Christian name has /ts/ (see Visser (2010:48-49)):

Full form	Pet name
Jetske	Jekke
Jitske	Jikke
Tsjitske	Tsjikke
Lutske	Lukke
Wytske	Wike (Wykke)
Martsen	Makke
Grytsje	Kike
Pytsje	Pike
Tsjeardsje	Keke / Kekke

From a historical point of view, /ts/ is secondary and /k/ is primary in (1) and (2) above. This seems to be the other way around in /ts/ and /k/ in the full form of names and their hypocoristic counterparts. A (gradual) change from /ts/ into /k/, however, is unlikely. It may safely be assumed that in the formation of hypocoristics the cluster /ts/ of the full form was simply replaced by /k/. In order for this to be possible, there must have been an identification of /ts/ with /k/. The disjunctive lexical representations of some verb stems — with final /k/ and /ts/, see below — may have helped paving the way for this.

A fourth argument can be based on the following set of verbs, all with an infinitive ending in /-a(:)jtsjə/ ( <a(a)itsje>), /-o:jtsjə/ (<oaitsje>), or /-{a/ɛ}jtsjə/ (<eitsje>):

These verbs display an alternation between [k] and [ts] in the paradigm: [ts] occurs in case the inflectional suffix is -je (infinitive; first person singular present tense; all plural persons present tense; imperative), [k] occurs in all other inflectional forms, in point of fact, before schwa (see paradigm of class II). The full paradigm of la(a)itsje to laugh is given in the table below:

Present tense	Past tense	Past participle	Imperative
ik la(a)itsje I laugh	ik lake I laughed	ik haw lake I have laughed	la(a)itsje laugh!
do lakest you laugh	do lakest you laughed
hy laket he laughs	hy lake he laughed
wy/jim/hja la(a)itsje we/you/they laugh	wy/jim/hja laken we/you/they laughed

The stem vowel also shows alternation, as becomes clear from the following overview:

Infinitive	2nd ps. sg. present tense	Plural past tense	Past participle
la(aitsje)	lakest [la:kəst]	laken [la:kən]	lake [la:kə]
koaitsje	kôkest [kɔ:kəst]	kôken [kɔ:kən]	kôke [kɔ:kə]
loaitsje	lôkest [lɔ:kəst]	lôken [lɔ:kən]	lôke [lɔ:kə]
ploaitsje	plôkest [plɔ:kəst]	plôken [plɔ:kən]	plôke [plɔ:kə]
meitsje	makkest [makəst]	makken [makən]	makke [makə]
smeitsje	smakkest [smakəst]	smakken [smakən]	smakke [smakə]
reitsje	rekkest [rɛkəst]	rekken [rɛkən]	rekke [rɛkə]
weitsje	wekkest [vɛkəst]	wekken [vɛkən]	wekke [vɛkə]

There is some system in these alternations: /ts/ is preceded by a long vowel + glide sequence, viz. /a:j/ or /o:j/, or by a falling diphthong, viz. /{a/ɛ}j/, which also ends in the glide – [j]; /k/, on the other hand, is preceded by a monophthong. The long vowel + glide sequence /a:j/ alternates with the long monophthong /a:/, whereas /o:j/ alternates with /ɔ:/. In the case of /a:j/ ~ /a:/ the long vowel remains constant; in the case of /o:j/ ~ /ɔ:/, there is a qualitative (height) difference, though /o:/ and /ɔ:/ share their backness specification. There are no simplex words ending in /-*o:jk/ and /-*ɔ:ts/, with which this verb alternation is consistent. The falling diphthong /{a/ɛ}j/ alternates with the short monophthong /a/ in case it is preceded by /m/, with /ɛ/ in the other cases (it is hard to see which connection there is between /m/ and /ɛ/). It is the head portion of the (complex) vocalic sequence that appears to be the constant factor here.

Although no more than eight verbs are involved, there is a true pattern here. All underived verbs with an infinitive ending in /-V(V)jtsjə/ have a paradigm in which /-k/- and /-ts-/ and the complex vocalic sequences and the monophthongs alternate in the way shown in paradigm of la(a)itsje to laugh and the stem vowel alternations above.

Originally, the stems of the verbs in (3) ended in /-k/, which allows for a straightforward comparison with the corresponding verbs in the other West Germanic dialects (cf. Dutch lach(en) /lɑx/ to laugh, kok(en) /kok/ to cook; to boil, plukk(en) /plʏk/ to pick, to pluck, mak(en) /mak/ to make, rak(en) /rak/ to hit, smak(en) /smak/ to taste, and wak(en) /ʋak/ to watch; loaitsje has a parallel in English look /lʊk/. Due to a process of palatalization and assibilation, stem-final /-k/ turned into [ts] when ending up before the palatal glide /j/ of the suffix -je. Besides, between the monophthong and /k/ (or /ts/) the glide /j/ developed.

One may wonder whether this is still the synchronic state of affairs in Frisian (see also Tiersma (1979:112-115) . That is to say, do the verbs at hand still have a stem-final /-k/, which turns into [ts] before -je? Or are there reasons to believe that nowadays the verbs at hand display stem allomorphy, so that the verb la(a)itsje for instance has the disjunctive lexical representation /l{a(:)jts/a:k}/, together with a statement concerning the distribution of the allomorphs?

A first argument for stem allomorphy is that the alternation between /k/ and [ts] is no longer productive. The lists of examples in (1), (2), and /ts/ and /k/ in full names and their hypocoristic counterparts form closed, non-extendable classes. This means that the alternation is unlikely to increase its scope.

A second argument is that some of the above verbs have undergone leveling, in two directions. In the first place, in some dialects the form with the monophthong followed by /k/ has spread throughout the paradigm, as a result of which la(a)itsje, koaitsje, loaitsje, ploaitsje, and weitsje have been replaced by laakje, kôkje, lôkje, plôkje, and wekje, as shown in the table below for laakje to laugh:

Present tense	Past tense	Past participle	Imperative
ik laakje I laugh	ik lake I laughed	ik haw lake I have laughed	laakje laugh!
do lakest you laugh	do lakest you laughed
hy laket he laughs	hy lake he laughed
wy/jim/hja laakje we/you/they laugh	wy/jim/hja laken we/you/they laughed

Secondly, in some dialects the verbs became members of a different weak class (see two classes of weak verbs), viz. from verbs of class II they turned into verbs of class I (see paradigm of class II and paradigm of class I, so that laaitsje for instance became laaitse, see the table below:

Present tense	Past tense	Past participle	Imperative
ik laaits I laugh	ik laaitste I laughed	ik haw laaitst I have laughed	laaits laugh!
do laaitst you laugh	do laaitstest you laughed
hy laaitst he laughs	hy laaitste he laughed
wy/jim/hja laaitse we/you/they laugh	wy/jim/hja laaitsten we/you/they laughed

As the tables above show, laakje and laaitse have fully regular paradigms — of the second and the first weak class, respectively —, so paradigm regularization might be viewed as one of the driving forces behind the changes at hand. However this may be, these regularizations are easier to understand if disjunctive lexical representations are assumed.

Furthermore, the table below lists nine strong/irregular verbs with a stem ending in /-ɛk/, which have a past participle ending in /-øtsən/ ( -utsen), a pattern which the verb strike /strikə/ to iron has joined:

berekk(e)	berutsen	to cover up (a fire) – covered up
brekk(e)	brutsen	to break – broken
dekk(e)	dutsen	to cover - covered
rekk(e)	rutsen	to stretch - stretched
sprekk(e)	sprutsen	to speak - spoken
stekk(e)	stutsen	to stab, to sting – stabbed, stung
strekk(e)	strutsen	to stretch - stretched
trekk(e)	trutsen	to make an extract (from) – made an extract (from)
wrekk(e)	wrutsen	to avenge - avenged
strik(e)	strutsen	to iron - ironed

Diachronically, /-øtsən/ has developed from /-ukin/, which might be considered as an indication that /ts/ is a single segment here.

Again, this is a true pattern. All strong/irregular verbs with a stem ending in /-ɛk/ have a paradigm in which /k/ and /ts/ alternate as in the table above. Analyzing this alternation as an instance of synchronic palatalization and assibilation would ran into the same problems as with the verbs in (3) above. Therefore, disjunctive lexical representations are also posited for the verbs at hand.

Stem tense regularization is an extra argument for such representations (see also Tiersma (1979:114-115)). Take the verb brekke to break, which had the principal parts brek /brɛk/ (present tense stem) – briek /briək/ (past tense stem) – bruts(en) /brøts/ (past participle). Many spreakers have replaced the original past tense stem briek by bruts (brekke – bruts – brutsen), which means that a verb with three stems has been reduced to a verb with only two. According to Tiersma, the past participle is unmarked in comparison with the past tense stem, which may explain the direction of this regularization. But again, the latter is easier to understand if disjunctive lexical representations are assumed.

All in all, none of the above cases of the /k/~/ts/ alternation call for a synchronic palatalization and assibilation process. Diachronically, however, this process is involved in the transition from /k/ to /ts/, which can be inferred from the fact that the change only took place if /k/ preceded a front vowel (see Lin (2011)). The first stage must have been that /k/ fronted as a result of co-articulation. The fronting proceeded so far, that at a point in time (original) /k/ became virtually indistinguishable from [t], after which it was reinterpreted as such. After that, /t/ was affricated. The synchronic facts, however, point to a re-analysis of /ts/ from affricate to obstruent cluster.

/s/ + plosive sequences behave as (two-rooted) complex segments in Frisian. If /ts/ is analyzed as such as well, Frisian has one homogeneous system of complex segments, which is to be preferred. So, although /ts/ is likely to once have functioned as an affricate, there are arguments that it no longer does so at present.

Word-initially, the cluster /ts/- has the distributional property that it must precede the vowel /i/ ( /i/, /i:/, /iə/, or the front glide /j/). This is expressed by the following morpheme structure constraint ('M' denotes 'morpheme'):

The /ts/ constraint is the synchronic reflex of a diachronic palatalization and assibilation process which affected [k] and [t] when they preceded a front vowel. As such, it is the price to be paid for the decision not to analyze initial /ts-/ as an affricate.

An indication that Frisian has one system of two-rooted complex segments is the fact that the order in which /s/ and the plosive appear in stems may have contrastive function, see the examples in the tables below:

tsien ten	stien stone
tsjitter(je) to twitter	stjitter(je) to stutter
tsjoer tether	stjoer steering wheel

guts gouge	gust barren, unserved
taats round-headed nail	taast touch; fingermark
lyts little	lyst list; frame

boks loud-speaker; (play)pen	bosk bundle; wood
fiks sturdy; firm	fisk fish
eks ex	esk ash

hyps blow, damage	hysp duck
hips thriving, healthy	hisp duck
wips flop, plop	wisp wisp (of straw)