1.1  Writing

As a reconstructed linguistic entity, of course, PS is unwritten. A brief overview of Semitic writing systems appears in Chapter 1, §3.

2.1  Consonants

Proto-Semitic is traditionally reconstructed with 29 consonants, all of which are preserved in the inscriptional Ancient South Arabian languages such as Sabaic (see Chapter 13). There is good evidence, however, for a 30th consonant, a glottalic velar (or uvular) fricative, *x’ (or *χ’), which merged with *x in East Semitic and with *ħ in West Semitic (Huehnergard 2003). As can be seen in Table 3.1, many of the consonants occur in triads of a voiceless, a voiced and a third member. The reflexes of the third members of the triads are pharyngealized or uvularized in Arabic (and the reflex of *k’ is the uvular q), but they are glottalic/ejective in the Ethiopian Semitic languages (see Chapters 6–10) and in Modern South Arabian languages such as Mehri (Chapter 11); there is also evidence that their reflexes were glottalic in some of the ancient languages, such as Akkadian and Hebrew (Cantineau 1951–52, Faber 1980, Steiner 1982, Kogan 2011a), and so it is likely that they were glottalic in the proto-language as well, and underwent pharyngealization in the history of Arabic (Zemánek 1996).

The PS triad of fricative laterals – *ɬ, *l and *(t)ɬ’ – is now well established (Steiner 1977; see also Voigt 1992). The PS consonants reconstructed as affricates are simple fricatives in most of the descendant languages, but, again, their affricated nature in the proto-language is suggested by features of Akkadian and Hebrew phonology (Steiner 1982; Faber 1985). The voiceless non-glottalic plosives were probably aspirated when syllable-initial. The fricative *s may have had a palatalized allophone in some environments, since its reflex is a palatal [ʃ] in several of the languages (Babylonian Akkadian, Hebrew, Aramaic, Jibbāli), rather than [s] as in Assyrian Akkadian, Arabic and Ethiopian Semitic. ³

The reflexes of the PS consonants in a representative sample of Semitic languages appear in Table 3.2.

All of the consonants could be geminated. This is also the case in some of the descendant languages, although in some, such as early Aramaic and Hebrew, the laryngeals and pharyngeals may not be geminated.

There is marked tendency for the consonants of Semitic verbal roots (§3.2) to remain phonologically consistent. Thus, conditioned sound changes that would affect only some forms of a root are often blocked by paradigmatic pressure, so that the root continues to exhibit the same consonant phonemes in all forms. Less frequent, but also well attested, is the opposite development, where a conditioned sound change spreads analogically to other forms of a root in which the conditioning factor is not operative (Huehnergard 2013). The following phonological processes involving the consonants may be observed in Proto-Semitic:

• *w assimilated to a following dental-alveolar stop: e.g., *ja-t-wabal (3-recp-carry.pcs) > *jawtabal (metathesis; see §3.5.5) > *jattabal ‘he carried along’. Reanalysis of the product of this change occasionally resulted in byform roots in some languages, e.g., both wabaːlum and tabaːlum ‘to carry’ in Akkadian (Huehnergard 2006a, 2014a). For other changes involving *w and the palatal glide *j, see under Vowels (§2.2).
• Regressive assimilatory voicing and devoicing of consonants occasionally resulted in byform roots and seemingly irregular correspondences among cognates. For example, the pcs forms (§3.5.2) of the PS roots *b-k’-r ‘to pierce, split’ and *p-k’-r ‘to want, need, claim’, viz., *jabk’ur and *japk’ur, could apparently be pronounced the same, [japk’ur]; reanalysis and the tendency for root integrity, noted above, yielded byforms, one with initial *b and one with initial *p, of both original roots, reflexes of which appear throughout the descendant languages (Huehnergard 2014b). ⁴

  Table 3.2   Reflexes of the Proto-Semitic Consonants in Some of the Main Descendant Languages; the Semitistic Representations Are Given in Parentheses

  Proto-Sem.	Akkadian (Babylonian)	Gəʕəz (Classical Ethiopic)	Mehri	Sabaic	Arabic (Classical)	Ugaritic	Hebrew (Biblical)	Aramaic (Syriac)
  ʔ	Ø/ʔ (Ø/ˀ)	ʔ (ˀ)	Ø	ʔ (ˀ)	ʔ (ˀ)	ʔ (ˀ)	ʔ (ˀ)	ʔ (ˀ)
  ʕ	Ø/ʔ (Ø/ˀ)	ʕ (ˤ)	ʕ/ʔ (ˤ/ˀ)	ʕ (ˤ)	ʕ (ˤ)	ʕ (ˤ)	ʕ (ˤ)	ʕ (ˤ)
  b	b (b)	b (b)	b (b)	b (b)	b (b)	b (b)	b (b)	b (b)
  d	d (d)	d (d)	d (d)	d (d)	d (d)	d (d)	d (d)	d (d)
  ð	ʣ/z (z)	z (z)	ð ( d )	ð ( d )	ð ( d )	d/ð (d/d )	z (z)	d (d)
  ʣ	ʣ/z (z)	z (z)	z (z)	z (z)	z (z)	z (z)	z (z)	z (z)
  g	g (g)	g (g)	g (g)	g (g)	ɟ (j,ǧ)	g (g)	g (g)	g (g)
  ɣ/ʁ	Ø/ʔ/x (Ø/ˀ/ḫ)	ʕ (ˤ)	ɣ (ġ/ǵ)	ʁ (ġ/ǵ)	ʁ (ġ/ǵ)	ɣ (ġ/ǵ)	ʕ (ˤ)	ʕ (ˤ)
  h	Ø/ʔ (Ø/ˀ)	h (h)	h (h)	h (h)	h (h)	h (h)	h (h)	h (h)
  ħ	Ø/ʔ (Ø/ˀ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)
  j	Ø/ʔ/j (Ø/ˀ/y)	j (y)	j (y)	j (y)	j (y)	j (y)	j (y)	j (y)
  k	k (k)	k (k)	k (k)	k (k)	k (k)	k (k)	k (k)	k (k)
  k’	k’ (q,ḳ)	k’ (q,ḳ)	k’ (ḳ)	q (q,ḳ)	q (q)	k’ (q,ḳ)	k’ (q,ḳ)	k’ (q,ḳ)
  l	l (l)	l (l)	l (l)	l (l)	l (l)	l (l)	l (l)	l (l)
  ɬ	ʃ (š)	ɬ>s (ś>s)	ɬ (ś)	? (š,s2 )	ç (š)	ʃ ? (š)	ɬ (ś)	s (s)
  (t)ɬ’	ʦ’/s’ (ṣ)	tɬ’> s’ (ḍ>ṣ)	ɬ’ (ź,ṣ́)	dˤ (ḍ)	(t)ɬˤ/(d)ɮˤ (ḍ)	ʦ’/s’? (ṣ)	ʦ’/s’ (ṣ)	ʕ (ˤ)
  m	m (m)	m (m)	m (m)	m (m)	m (m)	m (m)	m (m)	m (m)
  n	n (n)	n (n)	n (n)	n (n)	n (n)	n (n)	n (n)	n (n)
  p	p (p)	f (f)	f (f)	f (f)	f (f)	p (p)	p (p)	p (p)
  r	r (r)	r (r)	r (r)	r (r)	r (r)	r (r)	r (r)	r (r)
  s	ʃ (š)	s (s)	h (h)	ʃ (s,s1 )	s (s)	ʃ ? (š)	ʃ (š)	ʃ (š)
  t	t (t)	t (t)	t (t)	t (t)	t (t)	t (t)	t (t)	t (t)
  t’	t’ (ṭ)	t’ (ṭ)	t’ (ṭ)	tˤ (ṭ)	dˤ (ṭ)	t’ (ṭ)	t’ (ṭ)	t’ (ṭ)
  ʦ	ʦ/s (s)	s (s)	s (s)	s (ś,s3 )	s (s)	s? (s)	s (s)	s (s)
  ʦ’	ʦ’/s’ (ṣ)	s’ (ṣ)	ʦ’/s’ (ṣ)	sˤ (ṣ)	(t)sˤ (ṣ)	ʦ’/s’? (ṣ)	ʦ’/s’ (ṣ)	ʦ’/s’ (ṣ)
  θ	ʃ (š)	s (s)	θ ( t )	θ ( t )	θ ( t )	θ? ( t )	ʃ (š)	t (t)
  θ’	ʦ’/s’ (ṣ)	s’ (ṣ)	ð’ (ð̣)	ðˤ (ẓ)	ðˤ (ẓ)	θ’(ẓ)/ɣ(ġ/ǵ)	ʦ’/s’ (ṣ)	t’ (ṭ)
  w	w (w)	w (w)	w (w)	w (w)	w (w)	w(w)/j(y)	w(w)/j(y)	w(w)/j(y)
  x/χ	x (ḫ)	x (ḫ)	x (ḫ)	χ (ḫ)	x (ḫ)	x (ḫ)	ħ (ḥ)	ħ (ḥ)
  x’/χ’	x (ḫ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)	ħ (ḥ)

• After East Semitic broke away from the parent language, most of the remaining family underwent a change of prevocalic *s > *h in forms such as the 3rd-person pronouns, as in independent *siʔa > *hiʔa ‘she’ and suffixal *‑su > *‑hu ‘his’ (see §3.1.1); the adverbial ending *‑isa, as in *bajt-isa > *bajt-iha (house-dir) ‘to the house’ (§3.3.2.4, end); and the causative marker *s, as in *tu-sa-ʕlij > *tu-ha-ʕlij (2-caus-ascend.pcs) ‘you sent up’ (§3.5.5). This may be viewed as an incipient change affecting high-frequency, low-stress function words (and the close grammatical relatives of such words), which are known to undergo sound changes before other word classes do (Phillips 1983); ⁵ but further spread of the change was blocked in most verbal and nominal roots, again because of the pressure, noted earlier, for roots to remain consistent across paradigms. ⁶
• Assimilation of n to a following consonant is a regular feature of several of the languages, such as Akkadian, Ugaritic and Hebrew, and is also attested in later Sabaic and in the earliest Gəʕez inscriptions (but not in later Ethiopian Semitic); it may therefore have been an ancient dialectal feature or an areal phenomenon (Sanmartín 1995, Steiner 2012: 380–1).
• In Proto-Northwest Semitic, initial *w > *j: *warix- > *jarix- ‘month’; *waθab-ti > *jaθab-ti ‘you sat’ (sit.sc-2fsg).

2.2  Vowels

For PS a reconstruction of three short vowels, *a, *i, *u, and three corresponding long vowels, *aː, *iː, *uː, is uncontroversial. This system is preserved unchanged in Classical Arabic. In most of the other languages, various developments have obscured the original system to a greater or lesser extent. In Gəʕəz, for example, the two short high vowels merged to a central ǝ (IPA [ɨ]), as in bǝrk ‘knee’ < *birk‑, ʔǝzn ‘ear’ < *ʔuðn‑. In many dialects of Akkadian, a fourth vowel quality, short e and long eː, achieved phonemic status, as in Old Babylonian egrum ‘twisted’ < *ħagrum vs. igrum ‘hire’ < *ʔigrum vs. agrum ‘hired’ < *ʔagirum. In Hebrew, the short vowels were sometimes preserved, sometimes reduced, sometimes lowered or backed, depending on syllable structure and word stress. In most of the languages, the long vowels remained largely unchanged, although a diagnostic feature of Canaanite languages is the change of *aː to *oː.

There are no diphthongs in the usual sense of a sequence of two vowels (see the next section on syllable structure), but the sequences *aj and *aw are often referred to as diphthongs in Semitic studies, and often undergo simplification to long vowels (e.g., [eː] and [oː], respectively) in the descendant languages. The sequences *ij and *uw are generally equivalent phonetically to *iː and *uː, respectively, in the descendant languages and presumably so also in PS (thus, e.g., the Hebrew form rūm ‘height’ < *ruːm‑ ~ *ruwm‑ has the same historical pattern, C ₁ uC ₂ C ₃ , as ʕōmɛq ‘depth’ < *ʕumk’‑). The sequences *iw and *uj were unstable, also tending to become *iː and *uː, respectively (e.g., *t’uːb‑ ‘goodness’ < *t’ujb-), unless preserved by paradigmatic pressure.

The sequences VwV and VjV were sometimes unstable, tending to reduce to a single vowel. The following developments, for example, may be posited already for PS: ˈawa, ˈaja > aː/_CV, but aˈwa > u and aˈja > i/_CC, as in *ˈt’ajab-uː > *ˈt’aːb-uː (good-3mpl) ‘they (m) are good’ but *t’aˈjab-ta > *ˈt’ib-ta (good-2msg) ‘you (msg) are good’ (Huehnergard 2005: 176–8). It is likely that CwV ₁ and CjV ₁ > CV ₁ ː in PS (unless preserved by paradigmatic pressure), as in *ja-kwun-uː > *ja-kuːn-uː (3-stable.pcs-mpl) ‘they (m) became stable’, *ja-ɬjim-uː > *ja-ɬiːm-uː (3-set.pcs-mpl) ‘they (m) set’.

2.3  Syllable structure and stress

PS has only three syllable types: short CV, and long CVː and CVC. Thus, syllable-initial and syllable‑final consonant clusters are not permitted, nor are sequences of vowels. Long vowels do not occur in closed syllables; when a long vowel would arise in a closed syllable through some phonological process, it is shortened; for example, the 3msg form corresponding to 3mpl *jaɬiːmuː (see end of preceding paragraph) is *ja-ɬim (3-set.pcs) ‘he set’, with short i. The restrictions on syllable types are overridden in various ways in most of the descendant languages, although Classical Arabic, for example, preserves the original syllable structure to a large extent (exceptions being long vowels before geminated consonants, as in the participial form maːd.dun ‘extending.msg.nom’ < *maː.di.dun).

Classical Arabic and Akkadian exhibit essentially the same assignment of word stress, which may therefore also be posited for PS, and is non-phonemic: stress falls on the right-most long syllable other than the final syllable: *ˈwaː.θi.bum ‘sitting.msg.nom’, *waː.ˈθib.tum ‘sitting.fsg.nom’, *waː.θi.ˈbaː.tum ‘sitting.fpl.nom’. Words with no long syllables are stressed on the first syllable: *ˈʕa.pa.rum ‘dust.nom’. ⁷ Bound form nominals (“construct forms”; §3.3.2.3) were morphosyntactically proclitic to their dependents and thus unstressed; a PS rule of vowel syncope probably affected such unstressed forms, e.g., nonbound *ˈwa.ri.xum vs. bound *war.xu ‘month.nom’, though the effects of this rule are frequently diminished by analogical leveling (Steiner 2012). There is probably also a narrower PS rule of vowel syncope, a > ∅/aC ₁ _ C ₁ V, as in *k’alalum > *k’allum ‘small.nom’.

3.1  Pronouns

3.2  Nominal and verbal roots

Most nouns and verbs in PS exhibit nonconcatenative morphology; that is, they consist of roots comprising invariable sequences of consonants (called the radicals of a root), over which are laid patterns – templates – that furnish the morphology of words. The patterns may be simple vowel melodies, such as a … i, but they may also exhibit gemination of the second or third radical or a prefixal or suffixal element or a combination of these. Forms of the root *s-l-m ‘(to be) whole’ illustrate some of these possibilities:

• *salim-at-um ‘whole-f-nom’ (adjective)
• *salaːm-um ‘wholeness-nom’ (substantive; an infinitive pattern)
• *ti-slam-iː (2-whole.pcs-fsg) ‘you became whole’
• *nu-sallim (1pl-whole.fact.pcs) ‘we made whole’

The vast majority of verbs in PS are based on roots of three consonants, like *s-l-m. Internal reconstruction on the basis of PS forms, however, indicates that roots of two consonants occurred at an earlier stage. For example, some forms of certain roots with first radical w, such as *w-r-d ‘to descend’, lack the initial w, as in *ja-rid (3-descend.pcs) ‘he descended’ and the verbal noun *rid-at- (descend.inf-f) ‘descent’. ¹⁰ Further, some roots exist as byforms, with the third radical either a glide or a reduplication of the second radical, as in *r-b-j ~ *r-b-b ‘(to be) great’. Finally, some of the languages exhibit roots with two radicals reduplicated, as in Arabic z-l-z-l ‘to shake’. (Biradical roots are more common in other Afro-Asiatic languages.)

A few roots with four discrete radicals occur in most of the descendant languages, such as Akkadian b-l-k-t ‘to jump’, Gǝʕəz d-n-g-ṣ́ ‘to be dismayed’; as in these examples, the second radical is frequently a sonorant. Most such roots are restricted to a single language or subgroup, and so it is difficult to reconstruct any of them to the proto-language.

There are certain phonological constraints on the constituents of a root: while roots with identical second and third radicals are common, such as *m-d-d ‘to measure’, roots with identical first and second radicals cannot be reconstructed to the proto-language (rare examples are found in some languages, but they are the result of later developments; e.g., Gəʕəz s-s-l ‘to recede’ < *s-l-s-l), and roots with identical first and third radicals are rare. ¹¹ Further, homorganic consonants are generally not found as adjacent radicals (Greenberg 1950).

3.3  Nominals

Across Semitic, most adjectives, like *salim-at-um, are associated with verbal roots. Many substantives, too, like *salaːm-um, may be said to derive from verbal roots. The patterns of such adjectives and substantives are sometimes salient. C ₁ aC ₂ VC ₃ , for example, as in *salim‑, is a common verbal adjective that tends to be resultative: *naθ’ir‑ ‘guarded’ from *n-θ’-r ‘to guard’; *waθib‑ ‘seated’ from *w-θ-b ‘to sit’. The pattern C ₁ aC ₂ a:C ₃ , as in *salaːm-, is a common verbal substantive, used as an infinitive in several of the descendant languages. A listing of some of the reconstructible patterns is presented in §3.3.1. In the descendant languages, the semantic ranges of many patterns shifted, and some patterns were replaced by others, or merged; it is therefore often not possible to reconstruct whole deverbal noun forms to the proto-language with certainty, but rather only roots and patterns (Fox 2003: 68).

There are also, however, many substantives that are primary, not associated with a verbal root (although a root may be extracted from such substantives, to create a denominal verb), and not necessarily triradical. Unlike many deverbal nouns, primary nouns can be reconstructed to the proto-language in toto. Examples are parts of the body, such as *raʔs‑ ‘head’, *ʕajn‑ ‘eye’, *ʔanp‑ ‘nose’, *jad- ‘hand’; kinship terms, such as *ʔabw‑ ‘father’, *ʔimm‑ ‘mother’, *bin‑ ‘son’, *ʔaxw‑ ‘brother’; ¹² features of the physical world, such as *ʔarɬ’‑ ‘earth’, *ʔabn‑ ‘stone’, *nahar‑ ‘river’, *tihaːm‑ ‘sea’, *ʕiɬ’‑ ‘tree’, *daθʔ‑ ‘grass’, *jawm‑ ‘day’, *warix‑ ‘month’, *san-at- ‘year-fsg’; some color terms, such as *laban‑ ‘white’, *waruk’‑ ‘yellow-green’. Extensive lists are provided in Fox (1998) and Kogan (2011b).

3.4  Numerals

3.5  Verbs

3.6  Prepositions

A number of words that function as prepositions can be reconstructed to PS. Some of these are originally substantives, used adverbially as bound forms, as in *wist’-a bajt-im interior-acc.bnd house-gen ‘within the house’. A substantival origin of other prepositions, however, is not evident; some of these are simple CV forms, which were probably proclitic, such as *ka= ‘like’, while others are CVC forms, such as *ʔin ‘in’ (Voigt 1999). These invariably govern the gen as well: *ka=kalb-im (like=dog-gen) ‘like a dog’, *ʔin libb-i-ja (in heart-gen.bnd-1csg) ‘in my heart’. Several forms have an optional ending *‑aj, e.g., *wist’aj ‘in, with’. The prepositions *ʕal(aj) ‘on, against’ and *ʕad(aj) ‘up to, until’ are associated with verbal roots, respectively *ʕ-l-j ‘to go up’ and (West Semitic) *ʕ-d-w ‘to cross, traverse’; but whether the prepositions or the verbal roots are primary is uncertain.

Since prepositions are bound forms, and since bound forms can govern clauses (see §3.3.2.3), some prepositions are also common as subordinating conjunctions (with or without a relative marker): *ʕad(aj) ʔi-smaʕ-u (until 1csg-hear.pcs-sbrd) ‘until I heard’.

A list of probable PS prepositions follows; forms ending with =, such as *bi=, are usually proclitic.

• *ʔin ‘in’ in East Semitic, but in West Semitic of restricted occurrence, e.g., Gəʕəz ʔǝn-bälä ‘without’ (= Akkadian in(a) balu), and in a form extended with (f) *‑t, *ʔin-tV (also *ʔin-t-aj), meaning ‘at, via’ (Gəʕəz ʔǝntä) and ‘with’ (Babylonian Akkadian itti, Hebrew ʔɛ t ); note also Gəʕəz ʔǝn-zä (in-rel) ‘while’
• *ʕad(aj) ‘up to, until’; cf. the West Semitic verbal root *ʕ-d-w ‘to cross, traverse’
• *ʕal(aj) ‘on, against’; cf. the PS verbal root *ʕ-l-j ‘to go up’
• *bajn(-aj) between’ (lost in Akkadian, but present in Eblaite)
• *bal ‘without, non-’ (Pat-El 2013)
• *bi= West Semitic, ‘in, with’ (loc/ins)
• *ha= ‘to, for’; in West Semitic, found only in Modern South Arabian languages (e.g., Jibbāli he=ʃ ‘to him’) and in the Ancient South Arabian language Ḥaḍramitic; in East Semitic, it is extended with enclitic *=na, as *ha=na > Akkadian ana, Eblaite ʔa ₅ -na /hana/ (Tonietti 2013: 51)
• *ka= also *kiː (and *kaj ?) ‘like, as’
• *la= West Semitic ‘to, for’ (dat/dir); apparently lost in East Semitic, unless the preposition is the same as the asseverative particle *la=, for which see §3.10
• *min(V) ‘from’; in Ethiopian, Tigre has mɨn, but in Gəʕəz the form has become ʔǝm and ʔǝmǝnnä via an obscure set of developments; lost in Akkadian, but Eblaite has two distinct prepositions, min ‘in’ and minu ‘from’ (Tonietti 2013: 82–8)
• *sin or *ʦin ‘toward, at’? Only in Eblaite (and one early Akkadian text), where the spelling si-in indicates initial *s or *ɬ (Tonietti 2013: 90–3), and in Ancient South Arabian, where the usual writing s ³ n indicates initial *ʦ (in Minaic, Qatabanic, and early Sabaic; but later Sabaic texts have s ¹ n, with *s)
• *wist’(aj) or *wast’(aj) ‘in, at’; as noted above, derived from a substantive *wi/ast’‑ ‘interior’

As in many other languages, prepositional phrases, comprising a preposition and a bound-form substantive, are common; examples found across Semitic (albeit with varying lexemes) are {to face.bnd} = ‘toward’, {in hand.bnd} = ‘through the agency of’, {in middle.bnd} = ‘within’, {like mouth.bnd} ‘according to’.

3.7  Conjunctions

PS coordinating conjunctions are *wa ‘and’, *pa/ʔap ‘and then, and so’ (these two are proclitic in some languages) and *ʔaw ‘or’. For subordinating conjunctions, see §4.8.

3.8  Adverbs

Only a few true adverbs may be reconstructed, e.g., interrogative *mataj ‘when?’. Most words used as adverbs are demonstratives, substantives, and adjectives, often in the accusative case, e.g., *jawm-am(=ma) (day-acc(=top) ‘today’ or ‘daily’). For the adverbial endings *‑u(m) and *‑isa, see §3.3.2.4, end.

3.9  Negation markers

Pat-El (2012) and Sjörs (2018) plausibly reconstruct two PS negators, both of which are attested in East and West Semitic: *laː is the standard negator, while *ʔal is a marked form (both [+neg] and [+vol]) restricted originally to negating the short prefix conjugation *jV-C ₁ C ₂ VC ₃ when used with injunctive modal force, as in *ʔal ja-ðkur (neg 3-invoke.pcs) ‘may he not invoke’. This original system is preserved intact only in the Northwest Semitic languages.

3.10  Other particles

Two presentative particles, *han and *hal ‘here is …’, may be reconstructed to PS (Hasselbach 2007, Pat-El 2009).

An existential particle in Central Semitic is *jiθ‑ ‘there is/are’; the cognate in East Semitic, however, is a finite verb, *j-θ-w, which in Akkadian came to signify ‘to have’ (Bar-Asher Siegal 2011). A negative counterpart, *ʔajn‑ ‘there is/are not’, appears in Northwest Semitic, perhaps related to a negative ʔin in Arabic and Ethiopic. ²⁶

Proclitic *la= marks asseveration or affirmation, and could be prefixed to virtually any form, with the exception of imperatives (Huehnergard 1983); e.g., Ugaritic:

Asseverative *la= marks the short prefix conjugation verb specifically as injunctive: *la=jaðkur (la=3.invoke.pcs) ‘may he invoke’.

The particle *law introduces hypotheticals: *law jaðkur ‘would that he had invoked’.

Enclitic *=ma is a topicalizing particle:

In Akkadian, =ma also topicalized whole clauses, and became the most common clause connector:

4.1  Word order

The ancient West Semitic languages such as Biblical Hebrew, Old Aramaic, Classical Arabic and Gəʕəz are predominantly VSO. Akkadian prose, conversely, is SOV. But there are exceptions in Akkadian; for example, Old Assyrian has a few examples of SVO and even VSO (Kouwenberg 2017: 698–703); further, Akkadian names with verbal elements are often VS (e.g., i-šme – il-um 3-hear.pcs – god-nom ‘the god has heard’), and Akkadian poetry shows relatively free word order. Moreover, in Eblaite, which is also East Semitic, there are numerous instances of VSO clauses (along with other orders). Thus it is likely that Proto-Semitic was VSO; as is usually suggested, the change to SOV in Akkadian is undoubtedly due to prolonged contact with Sumerian, which is also SOV.

In most Semitic languages (including Akkadian), heads precede modifiers (adjectives, genitives, relative clauses). ²⁷

4.2  Verbless clauses

Verbless clauses are a common feature of most Semitic languages. Predicate nouns were probably marked with *‑a in Proto-Semitic (see n. 21); whether such clauses were S–P or P–S or both is uncertain: ²⁸

Also reconstructible to PS is the use of anaphoric (=3rd person) pronouns as copulas; the noun subject is then essentially extraposed:

Predicates may also be adverbial adjuncts: See §3.10 for an existential particle *jiθ‑.

4.3  Noun modification

Two types of genitive phrases may be reconstructed, in both of which the genitive follows its head noun. In the type called a “construct chain,” the head and its genitive are simply juxtaposed, but the head noun lacks the ending that marks nouns as nonbound, as in

The other construction employs the relative marker, itself a bound form in apposition to the (nonbound) head noun (see §3.1.3):

This second type became rare in some West Semitic languages, such as ancient Hebrew and Arabic (Pat-El 2010).

In both types of construction, a clause could stand in the position of the genitive noun. For example, either of the following was possible. The first type became less common in West Semitic.

or

Only the first type of construction was used for pronominal possession in PS:

• *ʔanθ-at-u-su
• woman-f-nom.bnd-3msg
• ‘his wife’.

4.4  Definite article

See §3.3.2.3.

4.5  Agreement

Rules of agreement in PS are difficult to reconstruct with confidence. In most Semitic languages, attributive adjectives agree with their head nouns in gender and number (and case, if applicable), though not necessarily in boundness. In some Ethiopian Semitic languages such as Gǝʕəz, however, the concord is less strict for inanimates (see Chapter 6). In Arabic, broken plurals (§3.3.2.2) of inanimates are construed with fsg adjectives and verbs. In Akkadian, agreement in the plural depends on the morphology of the head noun; e.g., Old Babylonian bīt-um labir-um (house-nom old.msg-nom) ‘old house’, but plural bīt-āt-um labir-āt-um ‘old houses’ (house-fpl-nom old-fpl-nom). This contrasts with a West Semitic agreement pattern such as that of Biblical Hebrew, where the gender of the noun in the singular determines the gender of the adjective in the plural: ʔărå̄y-ō t šōʔăḡ-īm ‘roaring lions’ (lion.m-fpl roaring-mpl) (see Huehnergard 2006c: 17).

Verbs agree strictly with their subjects in Akkadian. In Classical Arabic, verbs are singular when they precede plural subjects; sporadic instances of this are also found in Biblical Hebrew. In Gəʕəz, again, agreement of the verb with a f or pl inanimate subject is optional.

4.6  Negation

See §3.9.

4.7  Interrogative sentences

Interrogative sentences are marked variously in the descendant languages – e.g., intonation or stress in Akkadian, proclitic particles in Arabic and Hebrew, enclitic particles in Ethiopic – and so it is difficult to reconstruct the PS situation.

4.8  Subordination

A common Semitic subordinating conjunction is *kiː, with a wide semantic range, introducing temporal, causal, comparative and object clauses (‘when, because, as, that’). As noted in §3.6, certain prepositions also function as subordinating conjunctions, as do bound form nouns, such as *jawm-a (day-acc.bnd) ‘when’. Relative clauses occupy the slot of attributives, and are introduced either by the relative marker or by a bound form; see §4.3 for examples. The syntactic role of the head noun in the relative sentence may be filled by a resumptive pronoun:

In East Semitic, finite verbs in subordinate clauses, both relative and other types, are obligatorily marked with a final *‑u if the verb has no other ending, ²⁹ as in the following Old Assyrian examples:

Other subordination markers, ‑na and ‑ni, are also attested in Akkadian, both probably deriving from *‑na (with dissimilation to ‑ni after aː). It is likely that this feature, in which finite verbs are marked as nominalized, was inherited from PS (Hasselbach 2012), with allomorphs *‑u after consonants and *‑na after vowels. The feature is lost in Ethiopian Semitic and in Modern South Arabian. But in a diagnostic development that characterizes Central Semitic, the PS short prefix conjugation form with the subordination marker *‑u/*‑na, e.g., 3msg *jaðkur-u, 3mpl *jaðkuruː-na, was reanalyzed as a new marked imperfective form, a form that completely replaced the inherited PS form *jiðakkar (see Chapter 1, §2.2.3). ³⁰

For ‘if’, we may posit PS *sin(=ma) (> Akkadian šumma, Aramaic hin, Gəʕəz ʔǝmmä, Arabic ʔin). The apodosis of a conditional sentence could be introduced by the coordinating conjunction *wa ‘and’. Both protases and apodoses of conditional sentences exhibit a rather perplexing range of verb forms across the Semitic languages.