Proto-Eigə-Isthmus

Proto-Eigə-Isthmus
Period	5th millenium BP
Spoken in	Huyfárah
Total speakers	unknown
Writing system	none
Classification	Eigə-Isthmus languages
Typology
Basic word order	VO
Morphology	agglutinative
Alignment	NOM-ACC
Credits
Created by	Corumayas, with help from Cedh, Radius, and others

Proto-Eigə-Isthmus is the reconstructed ancestor of the Eigə-Isthmus languages. It was spoken in eastern Peilaš, probably in the region that later became Huyfárah, during the 5th millenium BP.

The reconstruction of Proto-Eigə-Isthmus is based primarily on comparison of Eastern and Forest Miwan, Ngauro loans found in Ndak Ta, and Faraghin and its ancestor Proto-Isthmus.

Phonology

Syllable structure

The Proto-Eigə-Isthmus syllable may be conceived in terms of the following structure:

1. an onset consisting of an initial consonant *C_i, which may be followed by an optional liquid or glide *(G) (one of *r *l *j *w);
2. a vocalic nucleus consisting of a vowel *V; and
3. an optional coda, consisting of a final consonant *(C_f), which may optionally be preceded by one of *s *n *r *l *j *w to form a coda cluster.

The syllable structure is thus *C_i(G)V((s,n,G)C_f).

There are of course some further restrictions on possible syllables; for example, the combinations *ji *wu do not occur (although *ij *uw do). For restrictions on onset and coda clusters, see the more detailed discussions below.

Many PEI roots consist of a single syllable with this structure. However, no attested EI language is as monosyllabic as this might suggest; compounding and affixation are highly productive derivational processes in all branches of the family, and have been for thousands of years. Indeed, the difficulties of reconstructing PEI arise in large part from the multiple layers of prefixes and suffixes that have attached themselves to roots over time, often causing changes in the root itself before wearing away and being replaced by new affixes.

A prevocalic glide *(G) sometimes acts as part of the vocalic nucleus rather than the onset. For example, in the changes leading to Proto-Isthmus, sequences of [*w + vowel] have distinct vocalic reflexes when they follow an initial consonant (e.g. PEI *Cwa > PIsth. *Co). Similarly, postvocalic glides form diphthongs with distinct reflexes in the daughter languages.

Onsets

Onset consonant inventory

The following onset consonants are reconstructed for Proto-Eigə-Isthmus:

Labial	Coronal		Dorsal	Glottal
*p *b	*t *d	*ts *dz	*k *g	*ʔ
	*θ *ð	*s *z		(*h)
*m	*n		*ŋ
*w	*r		*j
	*l

Plosives are reconstructed at three places of articulation: labial *p *b, coronal (possibly dental) *t *d, and dorsal (presumably velar) *k *g.

There are also coronal affricates *ts *dz, and two pairs of coronal fricatives: sibilant *s *z and non-sibilant (probably dental) *θ *ð. These sounds have tended to shift around in the descendant languages, making them considerably more difficult to reconstruct than the plosives. (Labial fricatives are not reconstructed for PEI, although many descendant languages have one or both of /f v/; the Miwan labial fricatives are mainly descended from *θ *ð, while Proto-Isthmus *f descends from *w word-initially and *θ elsewhere.)

Nasals are reconstructed at the same three places of articulation as plosives. The labial and coronal nasals *m *n are perhaps the easiest to reconstruct of the PEI onsets, as they remain unchanged in most of the descendant languages and are found in many clear cognates. By comparison, the evidence for the dorsal nasal *ŋ is extremely murky: it seems to be uncommon, and has merged with other phonemes in many languages.

Liquids *r *l and glides *j *w are reconstructed as well. Of these, *l survives unchanged in most of the descendant languages; the others are preserved most clearly in the Miwan branch. A number of Eigə-Isthmus languages do not contrast /r l/ with each other, having deleted *r or shifted it to something else; on the other hand, several Isthmus languages (including Faraghin and Doroh) have innovated new /r/ despite its absence in Proto-Isthmus.

A glottal stop *ʔ is reconstructed for what would otherwise be vowel-initial syllables; this simplifies the syllable structure by making the onset obligatory, and explains some developments in the daughter languages. Most notably, the final consonants of verbal roots regularly develop as codas, even though they're nearly always followed by a stem vowel; this makes sense if the stem vowel was preceded by an onset *ʔ. Further evidence may be found in the tonal systems of the Eigə Valley languages: it's been suggested that *ʔ induced glottalization or creaky voice on nearby vowels. It's also possible that the phonemic glottal stop found in some Eigə Valley languages is partly descended from PEI *ʔ.

Finally, a second glottal onset *h has been suggested to account for some words which unexpectedly show breathy voice in Eigə Valley languages. However, since it has no consonantal reflexes and only affects one branch of the family, it may make more sense to simply reconstruct this as a tonal distinction in Proto-Eigə Valley.

Voicing

A simple two-way voicing contrast is reconstructed for all PEI obstruents (plosives, affricates, and fricatives). In plosives and affricates this contrast seems to have been somewhat unstable in the early Eigə-Isthmus family; the regular correspondences in onset position are:

Proto-Isthmus	*p	*t	*k	*b	*d	*g	*ts	*dz
Proto-Eigə Valley	*b	*d	*g	*p	*t	*k	*ts	*ts

That is, Proto-Isthmus voiceless plosives correspond to Proto-Eigə Valley voiced plosives and vice versa, while Proto-Isthmus voiced and voiceless affricates both correspond to Proto-Eigə Valley voiceless ones.

The evidence (notably from the affricates *ts *dz, whose PIsth. reflexes are not predictable from the merged PEV form) suggests that Proto-Isthmus preserves the original voicing, and that Proto-Eigə Valley onsets underwent the following changes:

1. *b *d *dz *g > *bʱ *dʱ *dzʱ *gʱ > *pʰ *tʰ *tsʰ *kʰ
2. *p *t *k > *b *d *g
3. *pʰ *tʰ *tsʰ *kʰ > *p *t *ts *k

The murmured or breathy-voiced onsets *bʱ *dʱ *dzʱ *gʱ induced breathy voice on the following vocalic nucleus before being devoiced to aspirates. This may have been the first step in the process of tonogenesis among the Eigə Valley languages.

Fricatives did not undergo these changes. The voiced fricatives *ð *z are reconstructed where Proto-Isthmus *d *dz correspond to Miwan /v z/: e.g. Faraghin daradan and Miwan varja(:) 'dance', both from PEI *ða(r)j-ʔa.

Onset clusters

Onset consonants may form clusters with a following liquid or glide (*r *l *j *w). Miwan preserves more of these clusters than the other branches; the following table shows the clusters found in wordlists from three Miwan languages:

Forest Miwan		Old Eastern Miwan		Eastern Miwan
pr pl   bl tr tw dr dw kl kw gr gl gw   fl fw sl sw zj zw       rj lj		bj tr tj tw dr dj dw kr kj kw gj gw qw fr fl fj fw vr vw sj zw χj nj ŋw rj		pr pl pj     tr     fw zw       rj lj lw

(The entries in italics are attested only medially, and thus some of them might not represent original onset clusters.)

The set observed in the reconstructed lexicon for Proto-Isthmus is somewhat different:

Proto-Isthmus
*pl *pj *bl *tl *tj *dl *dzj *kl *kj *sj *nj

Comparing these two sets, we can see that Proto-Isthmus has entirely lost *r and *w; other interesting differences include PIsth.'s lack of clusters beginning in *f (because this corresponds to Miwan w), and its very frequent *tl *dl (which have simplified to d t in Miwan).

Based on this comparison, a probable set of onset clusters may be reconstructed for Proto-Eigə-Isthmus (those in parentheses are unattested in any reconstructable vocabulary so far):

Proto-Eigə-Isthmus
(*pr) *pl *pj *br *bl *bj *tr *tl *tj *tw *dr *dl *dj *dw (*tsl) (*tsj) *tsw (*dzl) *dzj (*dzw) *kr *kl *kj *kw *gr *gl *gj *gw *θr *θl *θj *θw *ðr (*ðl) (*ðj) *ðw *sl *sj *sw (*zl) *zj *zw (*mj) *nj (*nw) (*ŋj) *ŋw *rj (*rw) *lj *lw

A few phonotactic restrictions can be discerned here: onset clusters of [labial + *w] are forbidden, as are [sibilant (*ts *dz *s *z) + *r]; *r *l cannot cluster with each other, nor with nasals. On the other hand, the absence of *pr, *tsj, and *dzw is doubtless just an accident of our incomplete knowledge of these languages. In between these extremes, it's difficult to tell what the status of such unattested clusters as *tsl *dzl, *ðl *ðj, or *mj *ŋj is. Further work is needed here; it may also be that some of the clusters found only in one branch or the other are innovations.

Another question worth exploring is whether there are onset clusters *ʔr *ʔl *ʔj *ʔw: if so, they would contrast with plain *r *l *j *w, and would develop differently in some instances (for example *wa *ʔwa would become *fa *o respectively in Proto-Isthmus).

Proposed correspondences

The charts below show the development of onset consonants and clusters in three early daughters of Proto-Eigə-Isthmus: Proto-Isthmus, Ngauro, and Proto-Miwan. It will be updated as comparative work continues.

(NB: Correspondences in bold are attested in fairly certain cognate pairs; those in italics are postulated to exist but currently unattested. Entries in (parentheses) are not actually found in the relevant lexicon at all. Particularly uncertain correspondences are marked with a "?", whether attested or not. As-of-yet unknown reflexes are left blank.)

Onset consonants

PEI	PIsth.	Ngauro	PMiw
*p	*p	(b)	*b
*t	*t	d	*d
*ts	*ts	s	*t
*k	*k	(g)	*g
*b	*b	p	*p
*d	*d	t	*t
*dz	*dz	s	*t
*g	*g	k	*k
*θ	*t¹, *f²	(h)	*f
*s	*s	(ś)³	*s
*ð	*d	(∅)	*v
*z	*dz	z	*z
*m	*m	m	*m
*n	*n	n	*n
*ŋ	*nj¹, *jn²	?(ŋ)	*ŋ
*r	∅	(j)	*r
*l	*l	l	*l
*j	*j	z	*j
*w	*f¹, ∅²	(w)	*w
*ʔ	∅	(∅)⁴	∅⁴
(*h)	∅	(∅)⁴	∅⁴

1. Initial.

2. Medial.

3. Possibly pronounced [θ].

4. Presumably with tonal effects.

Onset clusters

PEI	PIsth.	Ngauro	PMiw
(*pr) *br	*ʈ *ɖ¹		(*br) *pr
*pl *bl	*pl *bl		*bl *pl
*pj *bj	*pj *bj		*bj *pj
*tr *dr	*ʈ *ɖ	dj (tj)	*dr *tr
*tl *dl	*tl *dl		*d *t
*tj *dj	*tj (*dj)		*dj *tj
*tw *dw	*t *d¹		*dw *tw
(*tsl *dzl)	(*tsl *dzl)
(*tsj) *dzj	(*tsj) *dzj		*tj *tj
*tsw (*dzw)	*ts *dz¹		*tw *tw
*kr *gr	*ʈ *ɖ		*gr *kr
*kl *gl	*kl (*gl)		*g(l) *kl
*kj *gj	*kj (*gj)		*gj *kj
*kw *gw	*k *g¹		*gw *kw
*θr *ðr	*ʈ² *ɖ		*fr *vr
*θl (*ðl)	*tl *dl		*fl (*vl)
*θj (*ðj)	*tj (*dj)		*fj (*vj)
*θw *ðw	*t *d¹		*fw *vw
*sl (*zl)	(?*sl *dzl)		*sl (*zl)
*sj *zj	*sj *dzj		*sj *zj
*sw *zw	*s *dz¹		*sw *zw
(*mj)	(*mj)		(*mj)
*nj	*nj	z	*nj
(*nw)	*n¹		(*nw)
(*ŋj)	?*nj		(*ŋj)
*ŋw	?*nj¹		*ŋw
*rj	?*j		*rj
(*rw)	?∅¹		(*rw)
*lj	(*lj)		*lj
*lw	*l¹		*lw

1. With w-coloring of the following vowel; see vowel correspondences below.

2. Medially > *f, possibly with w-coloring of the following vowel.

Vowels

Vowel inventory

The vowel systems of the Eigə-Isthmus family show a large amount of variation, even among closely related languages:

Faraghin		Feråjin
i eu oi e o a		i iː u uː e eː o oː æ æː ɒ ɒː

The Miwan languages tend to have particularly small inventories (which they make up for by having length and tone distinctions):

Forest Miwan		Old Eastern Miwan
i iː iw iːw u uː   a aː		i iː iw iːw u uː e eː a aː

Comparison of Ngauro with the reconstruction of Proto-Isthmus is particularly important for understanding the PEI vowel system.

Ngauro		Proto-Isthmus
i u e aü ai~ɛ au~o a		*i *ij *u *uj   *e *ej *o *oj   *a *aj

A simple 4-vowel system is reconstructed for PEI, but with a full inventory of diphthongs formed by following semivowels:

Proto-Eigə-Isthmus
*i *ij *iw *u *uj *uw *e *ej *ew *a *aj *aw

The open vowel *a and close vowels *i *u are quite easy to reconstruct, being unchanged in many of the descendant languages. The front mid vowel *e appears somewhat less frequently, and seems to occur only in roots, not in affixes.

All of the diphthongs have tended to merge with monophthongs in many descendant languages, again making them somewhat difficult to reconstruct. In particular, only Isthmus languages distinguish between *e and *aj or *i and *ej; and only Eigə Valley languages distinguish between *a and *aw, *e and *ew, or *u and *uw. (On the other hand, ablaut patterns in the verbal morphologies of the EI languages help quite a bit to clarify the picture.)

Proposed correspondences

Here Old Eastern Miwan and Forest Miwan are listed separately, as they seem to have simplified the vowel system in slightly different ways. Vowels with preceding *w- are included to show their distinctive reflexes in Proto-Isthmus. There are many points of uncertainty here; the bold, italics, and parentheses have the same meanings as in the consonant charts above.

PEI	PIsth.	Ngauro	OEMiw	FMiw
*a	*a	a	a	a
*e	*e	ai~ɛ	e	i
*i	*i	i	i	i
*u	*u	u	u	u
*aj	(*aj)	ai~ɛ	e	i
*ej	*ej	i	i	i
*ij	*ij	e		?a, u
*uj	*uj	aü¹	a	i
*aw	*a	au~o	u	u
*ew	*e	e	iw	iw
*iw	*(j)u	e	iw	iw
*uw	*u	?aü¹		?i
*wa	*o	(wa)	wa	wa
*we	*o	(wai~wɛ)	we	wi
*wi	*ji	wi	wi	wi

1. Borrowed as /au/ in Ndak Ta, but /oi/ in Faraghin; probably pronounced something like [œy] in early Ngauro.

Codas

Coda inventory

Most consonants may be found in coda position:

Labial	Coronal		Dorsal	Glottal
*p *b	*t *d	*ts *dz	*k *g	(*ʔ)
	*θ	*s		(*h)
*m	*n		*ŋ
(*w)	*r		(*j)
	*l

There is no evidence for a voicing contrast in coda fricatives. On the other hand, although *ŋ is not well attested in codas, it seems clear that it must have existed. Coda glottals *ʔ *h, like the glottal onsets, might be reflected in the tonal systems of the Eigə Valley languages; there is no other evidence for them.

Postvocalic *j *w form diphthongs, which are discussed with vowels in the previous section. Coda *ts *dz *s seem to have merged to *s in PEV; when followed by a consonant in Proto-Miwan, this usually becomes breathy voice on the preceding vowel (indicated by *ʱ in the correspondence charts). (However, Forest Miwan does contrast coda /s z/, so it may be that coda *dz sometimes remained voiced in that branch.)

Coda clusters, consisting of a final consonant (usually a stop or affricate) preceded by one of *s *n *r *l *j *w, are well attested. In this prefinal position, *n assimilates to the place of articulation of the following stop, giving *mp *mb *ŋk *ŋg. However, *s does not assimilate to the voicing of the following consonant: in clusters of *s + a voiced consonant, the voicelessness of *s is preserved. Again, *j and *w produce diphthongs which are discussed above. The other coda clusters behave differently in different daughter languages, with the prefinal consonant sometimes affecting the vowel (as in Proto-Miwan, where *sC clusters were one source of breathy-voiced vowels) and sometimes the final consonant (as in Proto-Isthmus, where *rC clusters became retroflex stops).

Codas were not affected by the changes that reversed the voicing of onset stops and affricates in Proto-Eigə Valley, so voicing in codas is normally consistent across the entire family.

Proposed correspondences

Coda consonants

(For coda *j *w, see the diphthongs under Vowels.)

PEI	PIsth.	Ngauro	PMiw
*p	*p	?(p)	*p
*t	*t	t	*t
*ts	*ts	s	*s, *ʱ
*k	*k	k	?*k
*b	*b	?(b)	?(*b)
*d	*d	d	?*d
*dz	*dz	?s	?*s, *ʱ
*g	*g	?(g)	*g
*θ	*f		*f
*s	*s	s	*s, *ʱ
*m	*m	m	*m
*n	*n	(n)	*n
*ŋ	*j	(*ŋ)	(*ŋ)
*r	?∅		*r
*l	*l		*l
(*ʔ)	∅		?*ˀ
(*h)	∅		?*ʱ

Coda clusters

(For *jC *wC, see the diphthongs under Vowels.)

PEI	PIsth.	Ngauro	PMiw
*sC	*sC	sC	*ʱC
*nC	*nC	nC	*nC
*rC	*ʈ, *ɖ, ?*C¹	?*jC	?*rC
*lC	*lC	?C	*lC

1. Plosives > *ʈ *ɖ. Medial *rp *rb trigger w-coloring of a following vowel; word-finally, they become *ʈu *ɖu.

Sound changes

These are partly tentative and definitely incomplete.

Proto-Eigə-Isthmus > Proto-Isthmus

• Fricative changes:
  • ð z > d dz
  • initial θ > t
  • remaining θ > f

• Retroflexion:
  • tr, kr, rt, rk > ʈ
  • dr, gr, rd, rg > ɖ
  • pr, rp > ʈw
  • br, rb > ɖw
  • rts rs > ʂ
  • rdz > r
  • fr, rf > ʂw
  • rn, rŋ > ɳ
  • rm > ɳw
  • rl > ɭ (if this cluster occurs)
• Retroflexion is blocked when the cluster is immediately followed by l, j, or w.

• initial r > ∅
• remaining r > w

• initial w > f

• geminate consonants simplify

• ʔ, (h) > ∅
• intervocalic w > ∅ (probably not including /Vj_V)

• Hiatus reduction: before any vowel, or after a stressed vowel
  • first, unstressed a, e > ∅
  • then unstressed i u > j w

• W-coloring:
  • wa, we > o
  • wi > ji
  • wo wu > o u

• iw > ju
• remaining coda w > ∅

• w > u when not adjacent to a vowel

• ŋ changes:
  • initial ŋ > (ɲ) > nj
  • intervocalic ŋ > (ɲ) > jn
  • coda ŋ > (ɲ) > j (except perhaps before k g)

• j > ∅ after a retroflex consonant, an onset Cl cluster, or another j

• i e > u o between a labial consonant and a following retroflex consonant

• ʂ ɳ ɭ > s n l

(For subsequent changes in the Isthmus family, see Proto-Isthmus.)

Proto-Eigə-Isthmus > Proto-Eigə Valley

• onset voicing shift (and rise of breathy voice on vowels)
  • b d dz g > bʱ dʱ dzʱ gʱ > pʰ tʰ tsʰ kʰ + breathy voice on following vowel
  • p t k > b d g
  • pʰ tʰ tsʰ kʰ > p t ts k
  • (this shift is blocked in medial clusters of two stops with the same voicing, e.g. PEI gasd~gasd ‘stream~PL’ > PEV kasd~ga(s)d > Ng. kasdgad > NT kasadgad)

(Since *ts doesn't become voiced, *ts and *dz merge: cf. EMiw tun 'red' < *dzusn-, timpi:za 'necklace' < *tsimp-.)

• coda ts dz > s z
• ? z > s /V_C

• vowel mergers: e ej ew > aj i iw (although maybe ej > ij in some environments)

• intervocalic ʔ h (if they both existed) induce phonation—creaky voice and breathy voice respectively—on adjacent syllables, then are deleted

• the preposition ʔum > a prefix m-, which then assimilates to the POA of a following stop (perhaps some other unstressed initial vowels are deleted too; but ʔas- seems to be preserved as Miwan as-)

• onset tl dl > tɬ dɮ > ts dz (I think it'd be fun to have some hints of the lateral affricate stage in the daughters)

(This change has to happen after the merger of original *ts *dz > *ts; cf. FMiw dimbal 'wife' < *tlujb-.)

Proto-Eigə Valley > Ngauro

(Not all these changes occur before the beginning of Ngauro writing.)

• *aj *aw > ai au in open syllables, > ɛ o in closed syllables
• *iw, *ij > e (maybe [ei])
• *uw, *uj > aü (maybe [œy])

• ? *ɛl > a before a coda consonant
• (other changes involving coda liquids)

• ? *s > θ in onsets
• *ts *dz > s z

• *nj > *ɲ > *j
• *j > z (at least initial)
• *r > j (at least /C_V)

• *m > n /_[+coronal]

Proto-Eigə Valley > Meshi

The order of these changes is not completely certain. However the /*kw *gw/ > /fw zw/ change must take place before ca. 2250 BP.

• *a̤j *a̤w > *əj *əw
• phonation contrast becomes non-phonemic
• *aj *aw > e o
• *ʔ > ∅ / (if this is still present in PEV)
• *ts *dz → t d / in onsets
• *ts *dz → s z / elsewhere
• *w > v / #_ , V_V
• *a *e *o > ∅ / _#
• nasals assimilate in place to a following obstruent
• C > ∅ / _# in most polysyllabic words; some clusters are simplified rather than deleted
• C > ∅ / C_#
• *e *o > i u
• *kw *gw > fw zw
• [+obstruent] > ∅ / [+nasal]_, [+obstruent]_
• t d s z > ʧ ʤ ʃ ʒ / _i
• *iw *uj > *ø > e / _C, _#
• *uj > *øj > ej / _V
• *ij *uw > *əj *əw > aj aw
• *θ *ð > s z
• *ŋ → g / V_V
• *ŋ → j / u_, a_
• *ŋ → Ø (elsewhere)
• *r *l > j w / C_
• tj dj sj zj > ʧ ʤ ʃ ʒ (persistent)

Proto-Eigə Valley > Proto-Miwan

• *ts *dz > *t *d (only occurs in onsets)
• *θ *ð > *f *v

• *aj *aw > *e *o
• *o > *u
• ? *uw > *uj (maybe)

• *s > *ʱ (breathy voice on preceding vowel) /V_C$

(This change may be blocked by morpheme boundaries in the dialect that led to Forest Miwan.)

Breathy voice is now phonemic and found in at least a third of Proto-Miwan roots.

• ? *t, *d > *s~z /_[+plosive] (voicing assimilates to the following consonant)

• tonogenesis

Proto-Miwan > Forest Miwan

• *e > i
• ? *ij *uj > i/a/u (depending on environment)
• *ŋ > n

Proto-Miwan > Old Eastern Miwan

• ? some instances of *k *g > q χ (possibly due to early Hitatc influence)
• intervocalic ʔ develops (maybe *wu *ji > ʔu ʔi; possible Hitatc influence here too)
• *z > *s in codas (if this hasn't happened already)

Old Eastern Miwan > Eastern Miwan

• loss of e (probably > i)
• ? u > o /_m (maybe not)
• loss of q χ ʔ (maybe χ > ʁ)
• loss of ŋ (mostly > n)
• loss of coda obstruents
• ? loss of b d g (maybe by lenition to v z ʁ; or by devoicing to p t k, with tonal effects on adjacent vowels; or maybe both: devoiced word-initially, lenited medially?)

Old Eastern Miwan > Southern Miwan

• r > ʀ
• ? χ > ʀ
• ? other uvulars merge into velars except adjacent to /a/
• ʔ > h
• v > w / #_
• l > w / _C, _#
• word-final obstruents become voiced and acquire a trailing short echo vowel, e.g. *dje:zuf > dje:zuvu (this last change is areal; Pirikõsu also inserted vowels after word-final consonants around 500 YP)
• sj > ʃ
• o develops from a variety of sources (< u adjacent to original uvulars, < a /Vw_, < am, etc.)

Morphosyntax

Verbal morphosyntax

There are two layers of morphology that can be securely reconstructed for PEI verbs, traditionally known as the grades and the stem vowel.

Aspect: the grades

Aspect in PEI was marked by a somewhat heterogeneous set of inflections applied directly to the verb stem; the resulting forms are called grades. In the descendant languages, the grades have produced ablaut patterns and even, in the case of Miwan, derivational tone.

Aside from the unmarked Zero-grade and the Reduplicated grade, the grades were marked by infixes in some verbs and suffixes in others (dividing the verbs into two conjugational classes). These infixes and suffixes were clearly allomorphs of one another, but were not necessarily identical in form. Both infixes and suffixes could be syllabic or non-syllabic, depending on the phonological shape of the stem to which they were attached (the nonsyllabic forms were apparently the result of a reduction of unstressed syllables sometime before the PEI stage).

In infixing verbs, the infixes were inserted immediately before a stem-final consonant. When the stem ended with a single consonant, the infix was normally non-syllabic and a coda cluster was formed. The behavior of infixes in stems that already ended with coda clusters was more complex; syllabic infixes were used here, and probably came between the two consonants of the stem coda, but the details are still being worked out.

In suffixing verbs, the suffix immediately followed the stem. When the stem ended in a consonant, the suffix was normally syllabic. Verbs whose stems ended in a vowel were always suffixing, with a non-syllabic suffix normally forming a coda consonant.

(The infixing verbs were apparently descended from forms in which another set of suffixes followed the aspect suffixes; verb forms marked with these secondary suffixes became lexicalized, leaving the aspect markers trapped inside the newly formed stems. The suffixing verbs were descended from verb forms without the secondary suffixes. The function of the lexicalized secondary suffixes can only be guessed at; one possibility is that they marked something like a telicity distinction.)

Many originally suffixing verbs (if their stems ended in a consonant) could become infixing by analogy. In this way, the infixing pattern gradually spread through the lexicon, leaving fewer suffixing verbs in the descendant languages.

The original system is further obscured by the incomplete preservation of the grades: no one EI language seems to preserve all of them, and they have tended to become derivational. As a result the precise forms and meanings of some of them are difficult to reconstruct, and even the existence of others remains uncertain.

The following grades can be reconstructed with some assurance:

Grade	Marking	Aspect
Zero-grade	unmarked	perfective
Reduplicated grade	partial reduplication (first *CV of stem)	iterative and/or intensive
S-grade	*-(u)s	resultative
N-grade	*-(i)n	imperfective or continuous
J-grade	*-i ~ *-j	inceptive, inchoative

In addition to those listed above, there may have been up to three more infixing-suffixing grades (the R-, L-, and W-grades).

Voice and valency: the stem vowel

PEI verbs were obligatorily marked for voice with a suffix traditionally called the stem vowel: this suffix, consisting of a vowel with a glottal stop onset, marked the verb as either active (*-ʔa), causative (*-ʔi), or detransitive (*-ʔu). These suffixes could also be combined with each other to recursively change the valency of verbs (*-ʔi increases valency, while *-ʔu decreases it); when combined the vowels were not separated by another glottal stop, but merged into diphthongs (*-ʔaj *-ʔaw *-ʔij *-ʔiw *-ʔuj *-ʔuw).

This system is easiest to see in Ngauro, where it produces chains of derived verbs; for example, kaima 'learn, know' → causative kaimi 'teach' → middle of causative kaime 'study' (< PEI *gajm-ʔiw or *gem-ʔiw). It's also reflected (in a more limited way) in Faraghin: most verbs end in -an (< *-ʔa + *ŋwaw), but causatives are derived with -oin (< *-ʔaj + *ŋwaw) and past participles with -od (< *-ʔu-di-ʔa or *-ʔu-di-ha).

Deverbalization

PEI had several ways of forming nouns from verbal roots, with at least one acting at each layer of verbal morphology. At the innermost layer, the S-Grade very frequently forms resultative nouns in both branches of the family.

At the next layer, there seems to be an old nominalizing suffix *-ga that could be added to the root in place of a stem vowel. The semantics of this suffix are somewhat obscure, and it seems to have become unproductive early on; nouns derived with it were already lexicalized in PEI. Examples include *njadz-ga 'war', and possibly *na-ga 'lord, chief'.

Finally, there was a newer deverbalizing suffix *-di which followed the stem vowel. This suffix was quite regular, and remained productive in many descendant languages into the historical period. It is reflected in the Faraghin past participles in -od and the Ngauro and OEMiw action nominalizer -ti. (The latter was borrowed from Ngauro into Proto-Ferogh or Early Faraghin, producing the Faraghin action nominalizer -č.)

Other verbal marking

Most other verbal markers seem to be restricted to one group or another of descendant languages. A few examples are discussed here.

Agreement in Faraghin

Faraghin verbs agree with their subjects; agreement is marked by suffixes that are clearly derived from the personal pronouns. For example, the citation form of Faraghin verbs is the third person singular, marked with the suffix -n which is clearly related to the 3SG pronoun ni (both < PIsth. *njo, < PEI *ŋwaw). These agreement suffixes apparently date to Proto-Isthmus; it's still unknown whether any Eigə Valley languages have subject agreement.

(It seems that third person agreement affixes are often grammaticalized later than first and second person ones, if at all. So we could easily posit an older, more fusional pattern for first and second person agreement, maybe dating back to before Proto-Isthmus.)

Verbal markers in Miwan

The following is a list of verbal affixes (both inflectional and derivational) that appear in attested Miwan words:

• *d- appears to form patientive nouns in OEMiw, cf. dre:χurfi: "speech-flow.together" ~ *re: "to speak", possibly saχudwa "flat-food" ~ *wa "to eat"
• *-ju (very productive in OEMiw) forms agentive nouns
• *sa:v- marks a reciprocal action in OEMiw sa:vre:χati "not speaking to each other"
• *-s might be an exhaustive plural, cf. OEMiw *re:sti "election" (lit. "everyone speaking").
• *-stu appears to form a cessative aspect, cf. FMiw/OEMiw gwastu (n. or v.) "(to) dump" ~ FMiw gwa: "to use", FMiw nidri:stu "to regret" ~ nidri: "to feel warm towards"
• *-χa(s) negative suffix (OEMiw)
• *-ʔi privative (with nouns) or non-potential (with verbs) suffix in OEMiw, cf. gwa:ʔiti "not important, useless" (becomes -hi in Southern)

In addition to this, Old Eastern Miwan had a hearsay marker tu:zi, which apparently was a free-standing particle (loaned into Naidda as tuze "I guess"). This appears to be an inflected verb meaning something like "(it) is told", from PEI *dlum "to tell".

Nominal morphosyntax

Number

There seem to be two ways of forming plurals in Eigə-Isthmus languages:

• A PEI suffix *-wa is the source of the plural marker in Faraghin (where it's usually realized as -a).

• Reduplication marks plurals in Ngauro (maybe with the additional meaning 'many', as in kasd~ga(s)d 'many streams'). It's possible that this is an innovation in the Eigə Valley branch.

De-nominalization

PEI had a suffix *-ʔa or *-ha that derived adjectives from nouns. This suffix was quite regular and remained productive in both branches of the family: it can be seen in Ngauro dyaka 'kingly, royal' (< PEI *trelk-ʔa or *trelk-ha), and in the Isthmus participles formed with Proto-Isthmus *-dja (< *-di-ʔa or *-di-ha).

Case and adpositions

Case prefixes

PEI had two morphemes that were either prepositions or case prefixes: *ʔas(-) and *za(-).

In Proto-Isthmus they are definitely case prefixes (genitive-accusative *as- and dative *dza-), and they still fill this role in the Faraghin pronoun system; some other Isthmus languages also retain an oblique noun stem reflecting the *as- prefix.

Elsewhere they seem to have become derivational:

• Several FMiw nouns and adjectives begin with as-, likely < PEI *ʔas(-). There are also a good number of Faraghin nouns that exhibit the consonant mutation caused by PIsth. *as-.

• It's also been proposed that the Meshi-derived word zafwi(:)ta(:) 'bird's nest' should be interpreted as < *za-gwenta 'LOC-bird', with a locative *za- < PEI *za(-). The reflexes of PIsth. *dza- also form locative adverbs in Doroh, and directional ones in archaic Boésin.

Other adpositions

There is at least one other morpheme that becomes a nominal affix in both branches of the family: the name Miw seems to be formed from PEI *pews with a nasal prefix derived from the same *ʔum that becomes a genitive case suffix in Faraghin and Feråjin. (The first syllable of Meshi is cognate with Miw, so this word—including the prefix—apparently goes back to Proto-Eigə Valley.) It seems that *ʔum was a preposition in the branch that led to the Eigə Valley languages, and a postposition in Isthmus.

Isthmus

The oldest case markers in the Isthmus languages are the two prefixes mentioned above. But newer case markers (at least in Faraghin, Feråjin, and Doroh) are suffixes derived from postpositions. This suggests that the language changed from prepositional to postpositional early on—certainly before the Western Isthmus stage. Since postpositions are pretty strongly (though not exclusively) associated with OV order, that order is the most likely for Proto-Isthmus; but the old prefixes hint at an earlier VO, head-initial stage of the language.

Miwan

FMiw di:za: "to wilt" might be formed from za: "to die" (attested in OEMiw) by prefixing a morpheme possibly meaning "away". If this is derived from an adposition, it would support the assumption that PEI was prepositional.

Compounding and noun-phrase syntax

Aside from this, there are some other similar developments in both branches, and analysis of noun-noun compounds has proved fruitful for understanding the changes to word order that have happened in the EI family.

Compounding in Faraghin

In Faraghin compounds, a genitive follows its head noun (NG), but adjectives precede (AN). (These compounds are probably not very old, since they include semantic fields like political, military, and nautical terminology that were probably developed after the Faraghin conquest of Huyfárah. On the other hand the compounding rule may be older than the rule used for clauses.)

In any case, NG order normally coincides with VO order; NG with OV is much rarer, especially when combined with AN (WALS, out of a sample of 1099 languages, lists only one—Tigré—with this combination of features). Since Faraghin compounds are AN, it seems that the NG order should be associated with an earlier VO order. So once again we have evidence for an early head-initial stage.

Compounding in Miwan

In Miwan, compounds with both head-initial (NA/NG) and head-final (AN/GN) order are attested. However, the distribution of these compound patterns gives a fairly clear picture:

• Old Eastern Miwan, the best-attested language of the family, is overwhelmingly head-final. The only known compound in OEMiw that might possibly be head-initial is dre:-χurfi: "speech-flow.together", which is reflected in Naidda jeorvi "chorus, harmony", but it seems rather likely that *χurfī is actually a nominalised verb form, which would make the compound one of the common GN type.
• Later forms of Eastern Miwan seem to also contain head-initial compounds, including NA compounds such as EMiw zuka-tun "fruit-red" > "tomato" and pa:n-tun "music-red" > "guitar". This might be an indication that these dialects switched their compound order at some point during the 1st millennium YP (possibly through influence from Edastean).
• Both genitive compounds attested in Late Eastern Miwan, fur-zin "tree-life" > "elm tree" and za-fwita: "nest-bird", follow NG order and, interestingly, have almost identical cognates in western Miwan: FMiw fur-zi:n and za-fwi:ta, respectively. If these are not borrowings, one can conclude that genitives followed the noun in Proto-Miwan or earlier, and that this ordering was changed later on at least in the eastern dialects. A strong argument in favour of this is presented by the word za-fwita:/-fwi:ta, whose second element is clearly cognate to FMiw kwintas "bird", but exhibits - in both dialects! - several sound changes not present in the uncompounded form, and therefore appears to be a very old formation. The lenition process turning *kw into fw seems similar to the Western Isthmus change of *p t k > f s x after the genitive preposition *as-, but the word can't be a loan from WI since it would have been *ɣonta there. It is probably a borrowing from proto-Meshi *fwinta, whose classical Meshi reflex is fwina.
• Forest Miwan does have a head-final AN compound, ti:-vur "funny-tree" > "hickory tree", but this does not prove that FMiw also switched to head-final compounding because it is identical to a Late Eastern word and might be a borrowing.

As a summary, it seems likely that PEV (pending evidence from Ngauro and/or Meshi) or a slightly later form of the language (i.e. Proto-Miwan) had head-initial compounds, with at least Old Eastern Miwan—but quite possibly all Miwan languages—later switching to head-final order. My guess is that this would have happened before -2000 YP because the strong superstratum influence of Ndak Ta (which is head-initial) would certainly not have encouraged this change.

Furthermore, if PEV was probably head-initial, and a head-initial ancestral stage underlies PIsth., then it seems almost certain that PEI was head-initial.

However, if we allow that the proto-languages need not have been entirely either head-initial or head-final, a slightly different scenario emerges which may provide a simpler explanation. The attested compounds in Forest Miwan show the same orders as Faraghin: NG and AN. It may be that both languages inherited these compound orders from PEI, and that only Eastern Miwan has changed its compound orders—first shifting from NG to GN (to become more consistently head-final), and then later shifting to NG and (mostly) NA under Edastean influence.

If this scenario is correct, PEI turns out to have had a somewhat mixed word-order pattern (but not a particularly rare one according to WALS): NG, VO, and prepositions, but AN.

Proposed roots

Main article: Proto-Eigə-Isthmus/Lexicon