►II.21.A.h. THE SEMIVOWELS w, y
The sounds w و and y ي are unstable.
α $Caw.C <> $Cay.C
Note: the sign $ means syllable boundary, the sign # means word boundary.
§I. The sequences $Caw.C and $Cay.C can alternate, e.g. ħaw.luṋ حَوْلٌ and ħay.luṋ حَيْلٌ “power;” baw.nuṋ بَوْنٌ and bay.nuṋ بَيْنٌ “distance;” ʔaw.ba.·tuṋ أَوْبَةٌ and ʔay.ba.·tuṋ أَيْبَةٌ “returning.” Classical dialects of eastern Arabia had $Caw.C where Classical dialects of western Arabia had $Cay.C, e.g. ħaw.θu حَوْثُ in the east corresponded to ħay.θu حَيْثُ “the place where” in the west, and la.haw.tu لَهَوْتُ in the east corresponded to la.hay.tu لَهَيْتُ “I was distracted” in the west. A similar alternation مُعَاْقَبَةٌ affected the sounds $Caw.wa and $Cay.ya, e.g. ṣaw.wȃ.muṋ صَوَّاْمٌ in the east corresponded to ṣay.yȃ.muṋ صَيَّاْمٌ “keen on fasting” in the west, naw.wȃ.muṋ نَوَّاْمٌ in the east corresponded to nay.yȃ.muṋ نَيَّاْمٌ “much sleeping (person)” in the west, and ṣaw.wȃ.ɣuṋ صَوَّاْغٌ in the east corresponded to ṣay.yȃ.ɣuṋ صَيَّاْغٌ “goldsmith” in the west (see Chaim Rabin (1951), “Ancient West-Arabian,” p. 129 §i).
There are about a hundred verb stems that could end with –aw in some dialects and –ay in others, e.g. ħa.kaw.tu حَكَوْتُ vs. ħa.kay.tu حَكَيْتُ “[I] narrated,” ma.šaw.tu مَشَوْتُ vs. ma.šay.tu مَشَيْتُ “[I] walked,” etc.. Those verbs were listed in a poem by ʔ·Ibnu Mȃlik ابن مالك which ʔ·As-Suyȗṭiyy السيوطي cited in المزهر في علوم اللغة وأنواعها . I will cite this poem in a later section of this book.
§II. These alternations were not limited to Arabic. The Akkadian language often had Cȗ.C < *Caw.C where West Semitic (and Proto-Semitic) had *Cay.C. In Sabaic inscriptions from the early Christian era the word for a political leader was written qwl /*qawl–/, but by the 5th-6th centuries AD this had changed to qyl /*qayl–/ (see Encyclopaedia of Islam, “Ḳayl”). The word *qawwām– seems to be attested in Aramaic only as qayyām “enduring.”
§III. Perhaps the Semitic languages of the east tended towards backing or rounding of the vowel a and this caused the change ay > aw. Conversely, the Semitic languages of the west may have tended towards fronting of a and this caused the change aw > ay (the evolution could have been ay > oy > ow and aw > ew > ey). There are indications of such a difference in the dialects of Classical Arabic, e.g. the word muṣħafuṋ مُصْحَفٌ in the east corresponded to miṣħafuṋ مِصْحَفٌ “codex” in the west, both are altered from maṣħafuṋ مَصْحَفٌ (which is according to ʔ·As-Suyȗṭiyy السيوطي in الإتقان في علوم القرآن a loanword from Ethiopic maṣħaf መጽሐፍ “book”). A similar relationship existed between qudwa·tuṋ قُدْوَةٌ in the east and qidwa·tuṋ قِدْوَةٌ “model” in the west, qubulaṋ قُبُلًاْ in the east and qibilaṋ قِبِلًاْ “face to face” in the west, qinya·tuṋ قِنْيَةٌ in the east and qunwa·tuṋ قُنْوَةٌ “flocks” in the west, etc. (see more examples in Chaim Rabin p. 101, §k; compare also §II.16.D.).
β $Caw.C > $CōC; $Cay.C > $CēC
§I. In the modern spoken dialects the Classical sounds aw and ay are commonly pronounced ō and ē (in the Maghreb they are pronounced ȗ and ȋ). For example, the Classical words yaw.muṋ يَوْمٌ “day” and bay.tuṋ بَيْتٌ “house” are mostly pronounced yōm and bēt in the eastern dialects.
It seems that the changes aw > ō and ay > ē happen mostly in the terminal syllables of words. They are often not found in syllables of the type Caw or Cay that are followed by other syllables. For example, they are not found in words such as maw.ǵȗd for Classical maw.ǵȗ.duṋ مَوْجُوْدٌ “found, present,” ʔay.bas for Classical ʔay.ba.su أَيْبَسُ “stiffer, daw.xān for Classical daw.xȃ.nuṋ دَوْخَاْنٌ “dazed,” ɣay.rān for Classical ɣay.rȃ.nu غَيْرَاْنُ “jealous,” daw.wār Classical for daw.wȃ.ruṋ دَوَّاْرٌ “rotating,” and ṭay.yār for Classical ṭay.yȃ.ruṋ طَيَّاْرٌ “flying.”
Sometimes there are variants. For example, some people from Aleppo (including myself) say ħay.wān for Classical ħa.ya.wȃ.nuṋ حَيَوَاْنٌ “animal.” However, other people from Aleppo pronounce the same word ħē.wān. In this case it seems that the word ħaywān was changed to ħēwān by analogy, not by real phonetic change.
Similarly, the Aleppine word ʕaw.ᵈže·t “a turn” (from the root ʕwǵ عوج ) has the same origin as the word ʕō.ᵈža “slanted” which is used in other Syrian dialects. The word ʕō.ᵈža was perhaps *ʕaw.ᵈža, but it was changed to ʕō.ᵈža by analogy with terminal syllables which have ō < aw.
In my opinion, it is possible that the changes aw > ō and ay > ē started in secondary syllables (§II.I3.B.) of the forms CawC and CayC. It is possible that these changes were originally meant to avoid doubly-closed syllables.
§II. In a few modern dialects the Classical sounds aw and ay may be pronounced ȃ. This is found in dialects of the coastal region of Syria, where the word yām corresponds to Classical yaw.muṋ يَوْمٌ “day” and bāt corresponds to Classical bay.tuṋ بَيْتٌ “house.” In these dialects it is obvious that the change aw/ay > ā works only in terminal, doubly-closed syllables, because of the existence of such forms as bay.ti for Classical bay.tȋ بَيْتِيْ “my house.”
It appears that the sound ā < aw/ay was formerly pronounced *ō/*ē. This can be inferred from the fact that the sounds aw and ay in those dialects can correspond to Classical ȃ as well as aw and ay. It seems that the sounds ō < *ā and ē < *ā merged in those dialects with *ō < *aw and *ē < *ay, and next they were affected by a variation similar to that seen in bāt < *bayt “house” vs. bay.ti “my house.” For example, the word faw.ra “mouse” was created from fōr < fār < faʔ.ruṋ فَأْرٌ “mouse” by analogy with pairs such as *yōm/yaw.ma “day/her day,” and the word ᵈžay.ᵈži “chicken” was created from ᵈžēᵈž < dᵈžāᵈž < da.ǵȃ.ǵuṋ دَجَاْجٌ “chicken(s)” by analogy with pairs such as *bēt/bay.ti “house/my house.” The same analogy has affected words of foreign origin, e.g. bayk “feudal lord” was somehow created from Turkic bēg (in Aleppo this word is pronounced bēg).
Thus, it seems that words such as yām and bāt were formerly pronounced *yōm and *bēt. The sounds ō < *ā and ē < *ā are rare in those dialects and they are found only sporadically. It is likely that those dialects have undergone relatively recent sound changes ō > ā and ē > ā.
The same phonetic development (aw > ō > ā and ay > ē > ā) apparently happened in modern Bedouin dialects of western Naǵd (ʕUtayba·t عُتَيْبَةُ), where there are forms such as bāt for bay.tuṋ بَيْتٌ “house,” wān for ʔayna أَيْنَ “where?,” ʕalā-k for ʕa.lay-ka عَلَيْكَ “on you,” mā fā-h šay < *mā fay-h šay for mȃ fiy-hi šay.ʔuṋ مَاْ فِيْهِ شَيْءٌ “there is nothing in it,” and ʔ·Aḷ.ḷāh yah.dā-k < *ʔ·Aḷ.ḷāh yah.day-k for ʔ·Al.lā.hu yah.diy-ka اللَّهُ يَهْدِيْكَ “God guides you.”
The same development is also found in modern Jewish speech of central Yemen (see Chaim Rabin p. 65, §e).
It appears that the same development happened in the Classical dialect of the Ħȃriθ bn Kaʕb اَلْحَاْرِثُ بْنُ كَعْبٍ who inhabited Naǵrȃn. It is reported that they said ʔilȃ-ka إِلَاْكَ for ʔilay-ka إِلَيْكَ “to you(masc. sing.)“ and ladȃ-ka لَدَاْكَ for laday-ka لَدَيْكَ “with you(masc. sing.).” It seems that in this dialect this shift was generalized by analogy to such words as ʕalȃ-hȃ عَلَاْهَاْ for ʕalay-hȃ عَلَيْهَاْ “on her” and ʕalȃ-kum عَلَاْكُمْ for ʕalay-kum عَلَيْكُمْ “on you(masc. plur.)” (see Chaim Rabin ibid.).
M. C. A. Macdonald (in The Cambridge Encyclopedia of the World’s Ancient Languages) mentioned something similar in Ancient North Arabian inscriptions from Dadān, the preposition ʕly /*ʕalay/ “on” was written ʕl /*ʕalā/ when pronominal suffixes were added to it. Since the Ancient North Arabian pronominal suffixes did not end with vowels (see Ernst A. Knauf in The Qurʾȃn in Context (2010), volume 6, p. 224), the Dadānitic orthography perhaps means that ay was changed to ā in doubly-closed syllables, just like in modern dialects of coastal Syria, western Naǵd, and Yemen.
§III. The sounds *aw and *ay are not written in any of the Ancient North Arabian languages. For example, the word maw.tuṋ مَوْتٌ “death” is written in Ṣafaitic mt (perhaps for *mōt), and the word ǵay.šuṋ جَيْشٌ “army” is written gs₂ (perhaps for *gēs₂). This could be an indication that etymological *aw and *ay were pronounced *ow and *ey in Proto-Arabic, whereas the open pronunciations aw and ay are a later development.
———————————————————————————
۞ SIDE NOTE
The Ancient North Arabian inscriptions used writing systems very similar to the “Epigraphic South Arabian,” which is known in Classical Arabic by the name ʔ·al-musnadu اَلْمُسْنَدُ (this name comes from Sabaic ms₃nd 
“inscription”). The use of the characters w 
and y
in the musnad system is relatively rare, and it is even rarer in the Ancient North Arabian writing systems. There is a debate among scholars over the interpretation of these characters. Some believe that these characters indicate the sounds w and y and occasionally and arbitrarily long vowels. Others believe that these characters indicate only the sounds w and y. Yet others (M. C. A. Macdonald) believe that the character w was used in Dadȃn to indicate the vowel –ȗ (but not –w) only in the terminal position of words, while the character ywas used in Dadȃn to indicate –y (but not –ȋ) only in the terminal position of words. In other word positions and in other ANA languages the sounds –w and –y and the long vowels were unwritten.
The Ancient Arabian writing systems, like the Canaanite/Phoenician writing system, are derived from the Ancient Egyptian uniliteral hieroglyphs. The Ancient Egyptian signs w 
and ı͗ 
were not used to indicate long vowels, and they appear to have always indicated the sounds w and y. For example, according to Antonio Loprieno (in Ancient Egyptian: A linguistic introduction (1995)), the Ancient Egyptian plural suffix <–w> was pronounced *–aw, the word <ḫprw> “from” was pronounced *χupraw, the word <msḏw> “enemy” was pronounced *masʸc̣iw, and the word <Ḥrw> “Hōrus” was pronounced *Ħāruw. A similar situation is found in the Phoenician alphabet, where the characters w
and y
always indicated the sounds w and y but never long vowels. In my opinion, there is no real need to assume that the Ancient Arabian writing systems worked on different principles. It is likely that the Ancient Arabian characters w and ywere used just like in the Ancient Egyptian and Phoenician systems, that is, they were used solely for indicating the sounds w and y wherever they existed.
In the musnad writings (of Yemen) the word for “day” was written sometimes ywm and sometimes ym, and the word for “house” was written sometimes byt and sometimes bt. This variation in spelling is found sometimes in one and the same text. Perhaps this variation has to do with terminal vowels that are not indicated in the written text? The word ywm might stand for *yaw.mV (or rather *yow.mV), while ym might stand for *yōm. Perhaps some of the scribes were inclined to use vernacular words that lacked case-vowels? Such inconsistent use of the case-vowels is known in Old Arabic texts written with the Nabataean alphabet.
Etymological aw and ay are almost never written in the Ancient North Arabian inscriptions. M. C. A. Macdonald believes that theses sounds could have existed in pronunciation without being written. Ahmad Al-Jallad (in An Outline of the Grammar of the Safaitic Inscriptions (2015), p. 47) pointed to Greek spellings with αυ (au) for Ancient North Arabian *aw, e.g. φυλῆς Χαυνηνῶν (phylē̃s Chaunēnō̃n) “[of] Chaunite tribe” for Ṣafaitic ʔl Kn, and δαυρα (daura) for Ṣafaitic dr “place.” The Greek diphthong αυ never contracted to ō (in Modern Greek it is pronounced av before vowels and voiced consonant, and af otherwise). Thus, the spelling Χαυν– cannot mean anything but *Kawn–. Similarly, the spelling δαυρ– (daur–) for Ṣafaitic dr “place” cannot mean anything but *dawr–. The Greek spelling with αυ is found also in the pre-Islamic papyri, e.g. Αυσω (Ausō) for the name ʔAwsuṋ أَوْسٌ . According to Al-Jallad, the Arabic sound *aw is always represented by αυ except in one word found in the Petra papyri, that is μοφαα (mophaa), which stands for *mawfaʕa·tuṋ مَوْفَعَةٌ* “elevated area” (in the Qȃmȗs والمَيْفَعَةُ الشَّرفُ من الأرضِ). This word was also written μαυφαα (mauphaa) and μουφαα (mouphaa). Perhaps the last spelling with ου (for *ow) was the closest to the actual pronunciation.
So it seems that etymological *aw was not really pronounced ō, at least not in all words. It is worth noting that in all the mentioned Greek spellings the sounds aw and ay are not in terminal syllables. The word Χαυνην– (Chaunēn–) does not exactly correspond to Ṣafaitic Kn but rather to a disyllabic *Kny “Chaunite.”
The Latin name Claudius was written in Ṣafaitic ʔqlds₁. This Latin name could not have been pronounced **Clōdius at the time of the Ṣafaitic inscription, because the Latin diphthong au was preserved until much later times (it still survives in the modern Romanian language). It seems that the sound *aw was left unwritten in Ancient North Arabian, which is unusual in Semitic orthography (or any other orthography). One possible explanation is that this writing standard was set by a dialect which had completely contracted the sound *aw, or rather *ow, to *ō. For example, it is possible that the Ṣafaitic Bedouins learned their alphabet from a dialect which did not preserve the sound *ow. In that dialect etymological *ow would have been left unwritten because it was pronounced *ō throughout. The Ṣafaitic Bedouins could have adopted the orthographic standard of that dialect even though they pronounced *ow in its original sound (which was only slightly different from *ō), at least in some word positions.
According to citations by Macdonald, a single verb stem was written in three different ways in Dadȃn: ʔdq–, hdq–, and hwdaq–. This appears to be a verb I=w that was perhaps pronounced *howdaq– > *hōdaq–. A similar variation existed in Sabaic, where the verb hwfy /*howfay–/ “[he] fulfilled” was sometimes written hfy /*hōfay–/. The fact that scribes in Dadān were able to transcribe the sound *ow when it existed is perhaps an indication that they were the creators (or among the creators) of the Ancient North Arabian script. In any case, they had better command of this script than those Ṣafaites who were unable to transcribe *ow.
Etymological *ay was rendered in Greek by αι (ai), ε (e) or η (ē). For example, Σαιφηνος (Saiphēnos) for Ṣafaitic h-Ṣ₂fy (in Classical Arabic اَلْضَّيْفِيُّ “the Ḍayfite”), Ιαιθεου (Iaitheou) for Ṣafaitic Yθʕ (in Classical Arabic it would be *Yayθaʕu يَيْثَعُ*; c.f. the cuneiform writings Ia-ta-ˀ, Ia-ta-a, and Ia-u-ti-ˀ for the name of a north Arabian vassal king appointed by Esarhaddon after his conquest of the country of A-ri-bi), Ζαιδος (Zaidos) for Ṣafaitic Zd (in Classical Arabic Zayduṋ زَيْدٌ ), Οβαιδος (Obaidos) and Οβεδου (Obedou) for Ṣafaitic ʕbd (in Classical Arabic ʕUbayduṋ عُبَيْدٌ ), Θεμαλλας (Themallas) for Ṣafaitic Tm-lh (in Classical Arabic Taymu ʔ·Aḷḷāhi تَيْمُ اَلْلَّهِ “servant of God”), Ονηνος (Onēnos) for Ṣafaitic Ħnn (in Classical Arabic Ħunaynuṋ حُنَيْنٌ ), and in the Petra papyri Λελα (Lela) for Layla·y لَيْلَىْ . Greek αι or Latin ae was unwritten in Ṣafaitic Qṣr for Καĩσαρ/Caesar. Macdonald noted that by the time of the Ṣafaitic inscriptions (roughly between the first century BC and the fourth century AD) the spellings αι, ε and η all indicated the same sound e. Therefore, it is not possible to draw any conclusions from these spellings regarding the pronunciation of Ṣafaitic *ay. However, Al-Jallad notes that etymological *i was commonly rendered by Greek ε and η but never by αι. Thus, it seems that αι was meant to render *ay or *ey. According to Al-Jallad (in Graeco-Arabica I: The Southern Levant (2015)), in the Nessana papyri the name Ζοναινος (Zonainos) (for *Ð̣unaynuṋ ظُنَيْنٌ*) was written by a single scribe in three different ways Ζονινος (Zoninos), Ζονειννος (Zoneinnos), and Ζονενος (Zonenos). It appears that this scribe was trying to write the name in a more accurate way phonetically than the common spelling Ζοναινος. By the time of the Nessana corpus (6th-7th centuries AD) the spellings ι and ει both should have indicated the sound i in Greek, but since the scribe wrote also Ζονενος (and since the name is commonly written Ζοναινος) it would seem that he was trying to render a sound lying between i and ay, which could have been *ey.
In conclusion, it is possible that etymological *aw and *ay were pronounced *ow and *ey already in the ancestor language of all the Arabic dialects. The sounds *ow and *ey probably contracted quickly to *ō and *ē in some dialects; this would explain why the ANA orthography did not indicate these sounds. The (relatively) open sounds aw and ay of the standard language and of some modern dialects are perhaps a secondary development from older *ow and *ey. This secondary development could have happened in eastern Arabian dialects where verbs such as qōma “[he] rose, stood up,” xēfa “[he] feared,” *ɣazō “[he] raided,” and banē “[he] built” were changed to qȃma, xȃfa, ɣazȃ, and banȃ of the standard language (vide infra §II.21.A.h.δ., §II.21.A.h.η., and §II.21.A.h.ι.).
———————————————————————————
γ $Caw$ > $Cȗ$; $Cay$ > $Cȋ$
An important phonetic change is *Caw > *Cow > Cuw = Cȗ and *Cay > *Cey > Ciy = Cȋ. This can be seen in pairs such as baw.nuṋ بَوْنٌ and buw.nuṋ بُوْنٌ “distance;” ʔay.nuṋ أَيْنٌ and ʔiy.nuṋ إِيْنٌ “time;” raw.ʕuṋ رَوْعٌ and ruw.ʕuṋ رُوْعٌ “awe;” ray.ʕuṋ رَيْعٌ and riy.ʕuṋ رِيْعٌ “mountain pass;” ray.ruṋ رَيْرٌ and riy.ruṋ رِيْرٌ “decayed.” In the modern dialect of Aleppo (and other Syrian dialects) there is kiyf for Classical kay.fa كَيْفَ “how,” and šiy for Classical šay.ʔuṋ شَيْءٌ “thing.”
———————————————————————————
۞ SIDE NOTE
In the Semitic grammar of Edward Lipiński (p. 212) similar examples are mentioned from Andalusian Arabic and Sabaic, but Lipiński considers the forms with the diphthongs to be secondary. When he talks about the endings <–w> and <–y> of written Sabaic (p. 364) he reasonably concludes that these represent the diphthongs *–aw and *–ay, yet he still considers the Sabaic endings *–aw and *–ay to be secondary, emerging from older *–ȗ and *–ȋ.
The sounds aw and ay are rather rare in Arabic compared to the long vowels ȗ and ȋ. In my opinion, it is likely that many aw and ay sounds were prehistorically changed to ȗ and ȋ. Comparison with Sabaic writings shows that e.g. the word ħiynuṋ حِيْنٌ “time” was originally *ħaynuṋ حَيْنٌ , because it was written ħyn in Sabaic. It is not surprising that the word ħiynuṋ حِيْنٌ had originally the vowel a, because this word is probably related to ʔaynuṋ أَيْنٌ “time,” *ʔayan– > ʔȃnuṋ آنٌ “time,” ʔawȃnuṋ أَوَاْنٌ “time,” ʔawnuṋ أَوْنٌ “slow walking, stop for rest,” hawnuṋ هَوْنٌ “slow walking,” and perhaps also ħawluṋ حَوْلٌ “year.” Moreover, there is probably an ancient relationship with *ʕawam– > ʕȃmuṋ عَاْمٌ “year,” yawmuṋ يَوْمٌ “day,” and *dawam– > dȃma دَاْمَ “lasted” (and perhaps also ʔuwamuṋ أُوَمٌ which is a broken plural adjective that qualifies nights لَيَالٍ أُوَمٌ ). All these words seem to originate from an ancient biconsonantal root *wam/*wan that was extended with prefixes that had the vowel a.
The changes aw > uw and ay > iy are easy to understand. These are simple assimilation of a vowel to a neighboring sound. However, the claimed changes ȗ > aw and ȋ > ay are not so easy to understand, especially in Semitic which is very hostile to diphthongs or diphthong-like sounds.
The changes aw > uw and ay > iy probably lie behind many divergent forms in Semitic languages. In Arabic these changes may explain e.g. why the eastern Arabs said ʔalla-ðȗna اللَّذُوْنَ whereas the western Arabs said ʔalla-ðȋna اللَّذِيْنَ “those who/which.” Probably the original forms of these words were *ʔalla-ðaw (in the east) and *ʔalla-ðay (in the west). When these changed to *ʔalla-ðuw/*ʔalla-ðiy they were extended by the suffix –na (perhaps by analogy with the masculine plural endings). In the oldest Sabaic texts the masculine plural relative pronoun was written ʔl (probably *ʔallā). In later texts the terminal *–ā was changed to *–aw/*–ay (per §II.19.C.d.), and this was secondarily inflicted for case, so that it became in the nominative case ʔlw /*ʔallaw/ and in the oblique case ʔly /*ʔallay/. The form *ʔallay is probably the source of the form ʔalliy which is used in the modern spoken dialects of Arabic.
In a similar way we may explain why some Semitic dialects have the singular relative pronoun *ðū and others have *ðī. Probably the original form was *ðā. This changed (per §II.19.C.d.) to *ðaw/*ðay (c.f. the Qatabanic masculine singular ðw /*ðaw/, the Sabaic masculine dual ðy /*ðayā/, and the Classical Arabic masculine dual ðawā ذَوَاْ ). When *ðaw/*ðay was changed to *ðuw/*ðiy, it was sometimes inflected for case, probably because the endings *–uw/*–iy sounded like the nominal case endings.
The Hebrew demonstratives zɛh זֶה “this” and ʔēllɛh אֵלֶּה “these” probably go back to *ðay זַי* and *ʔillay אִלַּי*. The vowel *ī is stable in Tiberian Hebrew and it does not change to anything else (c.f. the first person singular pronouns ʔănî אֲנִי and the word kî כִּי “that, for, when”). The segol in Tiberian Hebrew goes back either to *i or *ay. It is unlikely that the segol at the end of אֵלֶּה goes back to *i, because this is a terminal unstressed vowel. If it was originally short, it would have been lost by the 1st millennium BC (like the case vowels). Similarly, the form hallā-zɛh הַלָּזֶה shows that זֶה probably did not end with a short *i.
In a similar way we can relate the Arabic word kay كَيْ “so” to the Akkadian kī “as, like” and the Hebrew kî כִּי “that, for, when.” All these probably go back to *kay, which comes from the demonstrative *kā per §II.19.C.d..
It is usually supposed that the Proto-Semitic masculine plural construct nouns ended with the *–ū (in the nominative) and *–ī (in the oblique). This supposition is based solely on the (partial) evidence of Classical Arabic, but it goes against the evidence of Old Akkadian, Assyrian, Hebrew, Aramaic, Ancient South Arabian, and Ancient North Arabian. In all these languages the masculine plural construct nouns ended with –ay in the oblique case. Hebrew and Aramaic do not have the nominative case, and Akkadian writing does not distinguish between ō < *aw and ū, but the written suffix <–w> of the masculine plural construct nouns of Ancient South Arabian and Ancient North Arabian indicates the pronunciation *–aw in the nominative case (c.f. the Ancient Egyptian plural suffix <–w> which was pronounced *–aw according to Antonio Loprieno (1995), Ancient Egyptian: A linguistic introduction). Similarly, the suffix <–w> on the masculine plural verbs in Ancient South Arabian and Ancient North Arabian writings indicates *–aw. This suffix is attested in the modern Arabic dialects (e.g. katabaw for Classical katabȗ كَتَبُوْا “[they(masc.)] wrote” in the dialect of the Baħȃrina·t الْبَحَاْرِنَةُ in eastern Arabia).
In fact, the orthography of the Qurʔȃn indicates that the pronunciation *–aw for the masculine plural construct ending and the masculine plural verbal ending existed also in the Arabic of Mecca, because these endings are always written ـــوا (with an ʔalif after the wȃw). This writing should be compared to the writing of Qurʔȃnic words such as الْرِّبَوا “usury,” شُرَكَوا “partners(nom. plur.),” ضُعَفَوا “weak(nom. plur.),” etc. In all likelihood, the ending ـــوا was pronounced –ow < –aw (see §II.21.A.h.η.VI. below).
The endings –ȗ and –ȋ on the masculine construct nouns of Classical Arabic are likely innovations. They could have arisen by analogy with the absolute endings, or by phonetic change *–ow > –ȗ and *–ey > -ȋ. The same is true for the ending –ȗ of the masculine plural verb.
———————————————————————————-
δ $Ca.wV$ > $Cō$; $Ca.yV$ > $Cē$; $Ca.wVC$ > $CoC$; $Ca.yVC$ > $CeC$
§I. When unstressed open syllables $wV$ and $yV$ were preceded by a vowel a, the whole sequence contracted to ȃ, apparently through the intermediate stages ō and ē. For example, *ˈqa.wa.ma > *ˈqo.wo.ma > ˈqō.ma > ˈqȃ.ma قَاْمَ “[he] rose, stood up,” *ˈǵa.ya.ʔa > *ˈǵe.ye.ʔa > ˈǵē.ʔa > ˈǵȃ.ʔa جَاْءَ “[he] came,” *ˈxa.wi.fa > *ˈxa.yi.fa (§II.21.A.h.ζ.IV.) > *ˈxe.yi.fa > ˈxē.fa > ˈxȃ.fa خَاْفَ “[he] feared” (the forms qōma, ǵēʔa, and xēfa are attested in Sibawayh chapter 477 and in ʔIbn ᵈŽinnȋ, vol. I, p. 50; see also Chaim Rabin, p. 111, §y), *ˈṭa.wu.la > *ˈṭo.wu.la > *ˈṭō.la > ˈṭȃ.la طَاْلَ “[he] is/became long,” *ˈma.ya.hun > *ˈme.ye.hun > *ˈmē.hun > ˈmȃ.huṋ مَاْهٌ “water,” *ˈma.yi.hun > *ˈme.yi.hun > *ˈmē.hun > ˈmȃ.huṋ مَاْهٌ “watery,” *ħa.ˈma.wa.tun > *ħa.ˈmo.wo.tun > *ħa.ˈmō.tun > ħa.ˈmȃ.·tuṋ حَمَاْةٌ “female in-law,” and *fa.ˈta.ya.tun > *fa.ˈte.ye.tun > *fa.ˈtē.tun > fa.ˈtȃ.·tuṋ فَتَاْةٌ “young woman.”
§II. When unstressed closed syllables $wVC$ and $yVC$ were preceded by a vowel a, the whole sequence contracted to a short a, apparently through the intermediate stages *o < *ō and *e < *ē. The long vowels were shortened to avoid doubly closed superheavy syllables. Examples are *ˈda.ʕa.wat > **ˈda.ʕōt > *ˈda.ʕot > ˈda.ʕat دَعَتْ “[she] called,” *ˈra.ma.yat > **ˈra.mēt > *ˈra.met > ˈram.at رَمَتْ “[she] threw;” also in Form IV verbs such as **ʔa.ˈqōm.ta > *ʔa.ˈqom.ta > ʔa.ˈqam.ta أَقَمْتَ “[you(masc. sing.)] raised; dwelt” (this verb is derived from qȃma قَاْمَ “[he] rose, stood up”), and in nȗnated nouns of roots III=w/y such as *ˈʕa.ṣa.wun(nom.)/*ˈʕa.ṣa.wan(acc.)/*ˈʕa.ṣa.win(gen.) > **ˈʕa.ṣōn(nom./acc./gen.) > *ˈʕa.ṣon(nom./acc./gen.) > ˈʕa.ṣaṋ عَصًاْ “a stick, staff(nom./acc./gen.),” and *ˈfa.ta.yun(nom.)/*ˈfa.ta.yan(acc.)/*ˈfa.ta.yin(gen.) > **ˈfa.tēn(nom./acc./gen.) > *ˈfa.ten(nom./acc./gen.) > ˈfa.taṋ فَتًىْ “young man(nom./acc./gen.).”
§III. These changes are probably not very ancient, because in the Ṣafaitic Ancient North Arabian inscriptions we find forms such as myt for Classical *ˈma.wi.ta > *ˈma.yi.ta > *ˈme.yi.ta > ˈmē.ta > ˈmȃ.ta مَاْتَ “[he] died.” The Ṣafaitic spelling myt can mean neither **mēt nor **mȋt; this spelling perhaps means *meyita, since there is also an attested Ṣafaitic form mtt for Classical mētat > mȃtat مَاْتَتْ “[she] died.” The Ṣafaitic mtt perhaps means *mētat or *mȋtat. It is easy to imagine an alternation between *meyitat and *mētat or *mȋtat. Other examples from Ṣafaitic inscriptions are byt for bȃta بَاْتَ “[he] spent a night,” ṣyd for ṣȃda صَاْدَ “[he] hunted,” and ṣyr “[he] traveled.” According to M. C. A. Macdonald (in The Cambridge Encyclopedia of the World’s Ancient Languages), the verb *ˈka.wa.na > kȃna كَاْنَ “[he] is/was” was written in a Ṣafaitic inscription kn (this can stand for *kōna or *kūna).
ε $Ca.’wV$ > $Cȗ$; $Ca.’yV$ > $Cȋ$; $Ca.’wVC$ > $CuC$; $Ca.’yVC$ > $CiC$
§I. When stressed closed syllables $wVC$ and $yVC$ were preceded by an unstressed vowel a, the whole sequences contracted to u and i, apparently through the intermediate stages *ū and *ī. For example, *qa.ˈwam.ta > *qo.ˈwom.ta > **ˈqūm.ta > ˈqum.ta قُمْتَ “[you(masc. sing.)] rose, stood up,” *sa.ˈyar.ta > *se.ˈyer.ta > **ˈsīr.ta > ˈsir.ta سِرْتَ “[you(masc. sing.)] walked,” *ma.ˈwit.ta > *ma.ˈyit.ta > *me.ˈyit.ta > **ˈmīt.ta > ˈmit.ta مِتَّ “[you(masc. sing.)] died,” *ha.ˈyib.ta > *he.ˈyib.ta > **ˈhīb.ta > ˈhib.ta هِبْتَ “[you(masc. sing.)] feared,” *ṭa.ˈwul.ta > *ṭo.ˈwul.ta > **ˈṭūl.ta > ˈṭul.ta طُلْتَ “[you(masc. sing.)] are/became long.”
§II. According to Sibawayh (chapter 535), “some Arabs” said kȋda كِيْدَ and zȋla زِيْلَ for the standard verbs *ˈka.wi.da > *kēda > kȃda كَاْدَ “[he] almost [did]” and *ˈza.wi.la > *zēla > zȃla زَاْلَ “[he] ceased” (N.B. Sibawayh’s statement is quite clear, he means kȋda and zȋla, not kȇda and zȇla). The existence of forms such as kȋda and zȋla is not surprising, because it seems that some of the early dialects (particularly in the west) accented words of the form CV.CV.CV on the second syllable (see §II.15.). One would expect the Meccan dialect to have had forms such as kȋda and zȋla. In the Qurʔȃn such verbs are spelled with ʔalif زال، كاد . However, السجستاني says that he saw the verb ǵȃʔa جَاْءَ “[he] came” spelled with yȃʔ in Meccan codices. According to عاصم الْجَحْدَرِي , the verb ṭȃba طَاْبَ “[he] be(came) good” was spelled with yȃʔ in the model codex of ʕUθmȃn, and it can be seen spelled this way in the Kufic codex of Samarqand (which is the oldest manuscript of the Qurʔȃn). The Qurʔȃnic spelling with yȃʔ may mean the pronunciation ǵȇʔa and ṭȇba, which Sibawayh ascribed to “some of the people of the Ħiǵȃz” (وَهِيَ لُغَةٌ لِبَعْضِ أَهْلِ اَلْحِجَازِ). He specifically mentioned the Medinian poet كُثَيِّرُ عَزَّةَ as one of those who said ṣȇra for ṣȃra “reached; became.” However, given that there is no direct evidence for the Meccan pronunciation we cannot exclude that it was ǵȋʔa, ṭȋba, ṣȋra, etc.. Even if it was not, the pronunciation ǵȇʔa, ṭȇba, ṣȇra, etc. is not necessarily an original western Arabian pronunciation. As Chaim Rabin demonstrated in his study, the dialects of Medina and Mecca were not among the conservative ones in western Arabia. It could be that the original western Arabian pronunciation was ǵȋʔa, ṭȋba, ṣȋra, etc. but this was changed to ǵȇʔa, ṭȇba, ṣȇra, etc. under influence from eastern dialects. The dialects which, according to Sibawayh, said kȋda and zȋla could have been conservative dialects of western Arabia or Yemen.
A similar variation is seen in the Northwest Semitic languages, where Hebrew has mēṯ מֵת and Aramaic has mȋṯ ܡܺܝܬ for “[he] is dead.”
ζ $Ca.wVV$ & $Ca.yVV$
§I. In the standard form of Classical Arabic, non-terminal $Ca.wū$ is usually written Caʔȗ (perhaps for Ca.ȗ ? Vide infra §II.21.A.h.ο.III.). For example, *qa.wū.lun > qa.ʔȗ.luṋ قَؤُوْلٌ “much saying” and *ṣa.wū.lun > ṣa.ʔȗ.luṋ صَؤُوْلٌ “bully.”
§II. Non-terminal $Ca.yī$ becomes $Cay.yi. For example, *xa.yī.run > xay.yi.ruṋ خَيِّرٌ “good,” *ṭa.yī.bun > ṭay.yi.buṋ طَيِّبٌ “good,” and *ka.yī.sun > kay.yi.suṋ كَيِّسٌ “astute.”
§III. Non-terminal $Ca.wā$, $Ca.yā$ and $Ca.yū$ are stable. For example, ha.wȃ.ʔuṋ هَوَاْءٌ “air,” xa.yȃ.luṋ خَيَاْلٌ “ghost; fantasy,“ ha.yȗ.buṋ هَيُوْبٌ “much fearing,” ba.yȗ.ʕuṋ بَيُوْعٌ “much selling,” and qa.yȗ.ʔuṋ قَيُوْءٌ “much vomiting.”
§IV. Non-terminal $Ca.wi(i)$ often becomes $Ca.yi(i)$, and this then is affected by the changes described in §II.21.A.h.δ. and §II.21.A.h.ζ.II.. For example, *ˈma.wi.ta > *ˈma.yi.ta > ˈmȃ.ta مَاْتَ “[he] died,” *ma.wī.tun > *ma.yī.tun > may.yi.tuṋ مَيِّتٌ “dead,” *ha.wī.nun > *ha.yī.nun > hay.yi.nuṋ هَيِّنٌ “easy; contemptible,” and *na.wī.fun > *na.yī.fun > nay.yi.fuṋ نَيِّفٌ “exceeding.” Examples of unchanged $Ca.wi(i)$ are ṭa.wȋ.luṋ طَوِيْلٌ “long” and ʕa.wȋ.luṋ عَوِيْلٌ “wailing.”
η $Caw# > $Co#; $Cay# > $Ce#
§I. In the standard form of Classical Arabic, unstressed terminal –aw# and –ay# are changed to a sound called the “shortened ʔalif” الْأَلِفُ الْمَقْصُوْرَةُ (vide §II.12.). Originally this term seems to have meant a terminal short –a that was not dropped at pause like an original short –a. However, in Modern Standard Arabic the “shortened ʔalif” is usually pronounced a long –ȃ. In this book the letter ى is transliterated –·y, but in this section I am going to transcribe it –a.
Examples are, *ʔi.lay > ʔi.la إِلَىْ “to,” *ʕa.lay > ʕa.la عَلَىْ “on,” *ħat.tay > ħat.ta حَتَّىْ “until,” *ma.tay > ma.ta مَتَىْ “when,” *dun.ya.wu > *dun.yaw (§II.21.A.h.ι.III.) > dunya دُنْيَاْ “present world,” *ħublayu > *ħublay (§II.21.A.h.ι.IV.) > ħubla حُبْلَىْ “pregnant;” c.f. stressed terminal –aw# in law لَوْ “if” and stressed terminal –ay# in kay كَيْ “so that, in order to.”
§II. In some dialects of Classical Arabic unstressed terminal –aw# and –ay# had different pronunciations.
Sibawayh said (in chapter 477) that most of the etymological endings *–aw# and *–ay# have the sound of ʔalifu ʔ·al-ʔimȃla·ti أَلِفُ اَلْإِمَاْلَةِ “the ʔalif of inclination,” which usually means the sound ē, but in the word-terminal position it was probably a short e. There are only a few endings *–aw# that are not pronounced thus, e.g. *ʔ·al-qa.fa.wu > *ʔ·al-qa.faw > ʔ·al-qa.fa اَلْقَفَاْ “the nape” and *ʔ·al-ʕa.ṣa.wu > *ʔ·al-ʕa.ṣaw > ʔ·al-ʕa.ṣa اَلْعَصَاْ “the stick, staff.” The ending *–ay# in particles (such as *ħat.tay حَتَّىْ etc.) is not pronounced –e but –a. Sibawayh did not specify which dialects he was describing, but he excluded “many people of the Tamȋm and others” (وَجَمِيْعُ هَذَا لا يُمِيْلُهُ نَاسٌ كَثِيْرٌ مِنْ بَنِي تَمِيْمَ وَغَيْرِهِمْ). (See also Chaim Rabin, p. 115, §bb.)
§III. The orthography of the Qurʔȃn distinguishes strictly between etymological *–aw# and etymological *–ay#. In the Qurʔȃn etymological *–ay# is written ــى (a yȃʔ), but etymological *–aw# is written ـــا (an ʔalif). The Qurʔȃnic word *ʔ·ar-ri.baw الربوا “the usury” is an exception since it is always written with ـــوا (wȃw and ʔalif).
§IV. Sibawayh (in chapter 565 ↗) and other linguists (e.g. ʔ·Ibn Ǵinniyy اِبْنُ جِنِّي in سر صناعة الإعراب , vol. I, p. 50) stated that a sound ō existed in dialects of the Ħiǵȃz. They called this sound ʔalifu ʔ·at-tafxȋmi أَلِفُ اَلْتَّفْخِيْمِ “the ʔalif of amplification.” Sibawayh’s examples are ṣalȏ·tuṋ صَلَاْةٌ “prayer,” zakȏ·tuṋ زَكَاْةٌ “charity,” and ħayō·tuṋ حَيَاْةٌ “life.” He obviously picked those examples from the Qurʔȃn, where they are spelled حيوة، زكوة، صلوة . ʔ·Ibn Ǵinniyy mentioned the examples salȏmuṋ ʕalay-kum سَلَاْمٌ عَلَيْكُمْ “peace upon you” and qȏma Zayduṋ قَاْمَ زَيْدٌ “Zayd stood up.” Later authors (see Chaim Rabin, p. 105, §q) distinguished between two levels of tafxȋm, an “intense tafxȋm” تَفْخِيْمٌ شَدِيْدٌ (by which they meant ō) and a “modest tafxȋm” تَفْخِيْمٌ مُتَوَسِّطٌ (perhaps ɒ̄).
§V. The tafxȋm or back-rounded pronunciation of ā was an areal phenomenon of the 2nd millennium BC that covered a large area extending from Egypt to Syria. At that time the Proto-Arabic speakers are likely to have been located between Egypt and Palestine (in Sinai and the Eastern Desert of Egypt). It is likely that the shift ā > ō (commonly known as the “Canaanite Shift”) existed also in Proto-Arabic. When Arabic-speakers moved into Arabia in the Iron Age, they brought this shift as far south as Yemen (a back-rounded pronunciation of ā exists in the Modern South Arabian languages). In the modern dialects of Arabic a back pronunciation of ā exists in eastern Arabia (especially in Bahrain) and in western Syria (in the Alawite/Nuṣayri Mountains and in ʕAkkȃr عَكَّاْرُ the vowel ā may be sounded a full ō; in Homs and Damascus it may be sounded ɒ̄).
§VI. The ending ـــوا is sounded –ȗ in standard Classical Arabic, but in my opinion this is not how it was sounded in Mecca. The fact that it was used in the word الربوا indicates that it sounded close toـــا . Given that Meccan ـــا could represent a back-rounded vowel ȏ, we may assume that ـــوا represented a sound close to that.
It seems that the ligature ـــوا was originally meant to represent a sound lying between ـــو and ـــا , that is, between –aw and –ȏ. Perhaps this sound was –ow. When the ending was shortened to –o it became possible to write it with ـــا .
By analogy, we may presume that terminal ـى was originally used to represent the ending *–ey, and after this was shortened the same spelling represented –e (the ʔalif ʔ·al-ʔimȃla·t of Sibawayh and ʔ·Ibn Ǵinniyy).
§VII. According to Sibawayh (chapter 499), some groups in western Naǵd preserved the original pronunciation of the ending –ay at pause. They said ʔafʕay for ʔafʕa أَفْعَىْ “snake,” ħublay for ħubla حُبْلَىْ “pregnant,” and muθannay for muθanna مُثَنَّىْ “dual.” The Ṭayyiʔ طَيِّئٌ of northwestern Naǵd pronounced –aw and –ay even at junction.
وَذَلِكَ قَوْلُ بَعْضِ الْعَرَبِ فِي أَفْعَىْ “هَذِهِ أَفْعَيْ” وَفِي حُبْلَىْ “هَذِهِ حُبْلَيْ” وَفِي مُثَنًّىْ “هَذَا مُثَنَّيْ”. فَإِذَا وَصَلْتَ صَيَّرْتَهَا أَلِفًا. وَكَذَلِكَ كُلُّ أَلِفٍ فِي آخِرِ الْاِسْمِ. حَدَّثَنَا الْخَلِيلُ [الْفَرَاهِيدِيُّ] وَأَبُو الْخَطَّابِ [الْأَخْفَشُ] أَنَّهَا لُغَةٌ لِفَزَارَةٍ وَنَاسٍ مِنْ قَيْسٍ، وَهِيَ قَلِيلَةٌ، فَأَمَّا الْأَكْثَرُ الْأَعْرَفُ فَأَنْ تدَعَ الْأَلِفَ فِي الْوَقْفِ عَلَى حَالِهَا وَلَا تُبْدِلَهَا يَاءً، وَإِذَا وَصَلْتَ اِسْتَوَتْ اللُّغَتَانِ […] وَأَمَّا طَيِّئٌ فَزَعَمُوا أَنَّهُمْ يَدَعُونَهَا فِي الْوَصْلِ عَلَى حَالِهَا فِي الْوَقْفِ […] حَدَّثَّنَا بِذَلِكَ أَبُو الْخَطَّابِ وَغَيْرُهُ مِنْ الْعَرَبِ، وَزَعَمُوا أَنَّ بَعْضَ طَيِّئٍ يَقُولُ “أَفْعَوْ”. [كتاب سيبويه، باب الحرف الذي تبدل مكانه في الوقف حرفا أبين منه]
———————————————————————————
۞ SIDE NOTE
The dialect of Ṭayyiʔ طَيِّئٌ seems to have been a rather conservative dialect. This dialect preserved –aw and –ay in all positions. It also preserved the ending –at at pause (see Chaim Rabin p. 205, §y). It preserved the use of the feminine demonstrative tȃ تَاْ (the antiquity of which is attested in the Namȃra·t inscription). It preserved an archaic relative pronoun ðȗ ذُوْ , and it preserved an archaic conjugation of verbs III=w/y. It is unfortunate that this dialect was not described in detail by the linguists. They only commented on salient features that distinguished the poetry of this rather archaic tribe (this is the oldest attested tribe among the tribes of the Ǵȃhiliyya·t اَلْجَاْهِلَيَّةُ ).
———————————————————————————
§VIII. In my opinion, the available information is sufficient to draw a picture of how the endings –aw and –ay evolved in Arabic. It seems that these endings evolved just like the terminal long vowels (as described in §II.19.). Originally they were sounded –aw and –ay both at junction and pause. It seems that the exact pronunciation was –ow and –ey (§II.21.A.h.β.III.). In some dialects these endings were shortened at junction to –o and –e (for the same reason mentioned in §II.19.A.d., which is the avoidance of superheavy syllables of the forms CVwC and CVyC). These dialects are meant in Sibawayh’s statement that some people pronounced –ay at junction but ʔalif at pause. The ʔalif which he means is the “shortened ʔalif” الْأَلِفُ الْمَقْصُوْرَةُ which was probably a short –a. In Mecca this was pronounced –e (a shortened form of –ey). The last step (which happened in Mecca and in most other dialects, based on Sibawayh) was the pausal shortening of –ow and –ey which turned them to –o and –e in all positions, except when they had pronominal suffixes attached to them, e.g. namow-tu نَمَوْتُ “[I] grew,” bakey-tu بَكَيْتُ “[I] wept,” ʔiley-ka إِلَيْكَ “to you(masc. sing.),” and ʕaley-ka عَلَيْكَ “on you(masc. sing.).”
θ $Ca.wū# > $Caw#; $Ca.yī# > $Cay#; $Ca.wī# > $Ca.yī#; $Ca.yū# > $Ca.wū#
§I. Terminal *–a.wū and *–a.yī contracted to –awu and –ayi and then to –aw and –ay. For example, *ɣa.za.wū > ɣa.za.wu > ɣa.za.w غَزَوْا “[they(masc. plur.)] raided,” *yar.ḍa.wū > yar.ḍa.wu > yar.ḍaw يَرْضَوْا “[they(masc.)] are/become pleased(irrealis),” and *tan.sa.yī > tan.sa.yi > tan.say تَنْسَيْ “[you(fem. sing.)] forget(irrealis)” (the forms ɣazawu, yarḍawu, and tansayi are used in Classical Arabic when there is need for anaptyxis, see §II.16.C.).
§II. Terminal *–a.wī and *–a.yū changed to *–a.yī and *–a.wū and then they developed like these. For example, *tar.ḍa.wī > *tar.ḍa.yī > tar.ḍa.yi > tar.ḍay تَرْضَيْ “[you(fem. sing.)] are/become pleased(irrealis),” *ba.ka.yū > *ba.ka.wū > ba.ka.wu > ba.kaw بَكَوْا “[they(masc. plur.)] wept,” and *yan.sa.yū > *yan.sa.wū > yan.sa.wu > yan.saw يَنْسَوْا “[they(masc. plur.)] forget(irrealis)” (the forms tarḍayi, bakawu, and yansawu are used in Classical Arabic when there is need for anaptyxis, see §II.16.C.).
§III. Terminal –a.wā# and –a.yā# are stable. For example, ða.wȃ ذَوَاْ “those of(dual),” na.ma.wȃ نَمَوَاْ “[they(masc. dual)] grew” and ra.ma.yȃ رَمَيَاْ “[they(masc. dual)] threw.”
ι $Ca.wV# > $Co#; $Ca.yV# > $Ce#
§I. Terminal –a.wV# and –a.yV# were reduced in standard Classical Arabic to a “shortened ʔalif.“
In the Ṣafaitic inscriptions verbs III=w/y are written with terminal <–w> and <–y>, e.g. s₂tw for Classical šatȃ شَتَاْ “[he] be(came) in winter,” ʔty for ʔata·y أَتَىْ “[he] came,” bny for bana·y بَنَىْ “[he] built,” rʕy for raʕa·y رَعَىْ “[he] pastured,” and ʔʕly for ʔaʕla·y أَعْلَىْ “[he] raised up.” The Ṣafaitic spelling with terminal <–w> and <–y> probably means that terminal *–aw and *–ay (or rather *–ow and *–ey) were preserved. It is interesting that these sounds were only written in the terminal position of verbs. This is reminiscent of the situation found in Ethiopic, where *aw and *ay contracted to ō and ē in almost all positions except the terminal position of verbal forms, e.g. yətraxxaw (or yətraxxo) “[he/it] will open,” yəʔkay “[that he] be evil,” yəʕbay “[that he] be great,” bəkay “weeping,” ʔəkay “evil,” ʕəbay “greatness,” etc.
A general rule of standard CA is that terminal –u and –i fall off more easily than terminal –a. For example, the nominative case ending –u and the genitive case ending –i fell at pause, but the accusative case ending –a did not fall, the nominative and genitive case endings fell from nominal stems ending with –iy, but the accusative case ending –a did not fall (vide infra §II.21.A.h.ν.), the indicative mood ending –u fell from imperfective verb stems ending with –uw and –iy, but the subjunctive mood ending –a did not fall (vide infra §II.21.A.h.ν.).
A similar situation is found in Ethiopic, where terminal *–i and *–u disappeared completely, but terminal –a remained, and it appears even in verbs III=w/y, e.g. ʔatawa “[he] came,” ʕadawa “[he] transgressed,” ʔabaya “[he] refused,” bakaya “[he] wept,” samaya “[he] named,” etc.
Ahmad Al-Jallad (in An Outline of the Grammar of the Safaitic Inscriptions (2015), p. 49) suggests that in Ṣafaitic terminal *–i and *–u fell off from words, but terminal *–a remained, even in verbs III=w/y. He cites the Greek spelling αθαοα (athaoa) for Ṣafaitic *ʔatawa “[he] came” and βαϰλα (bakla) for Ṣafaitic *baql– “fresh herbage” with an accusative case ending. He also cites the Ṣafaitic rṣ₂y “[he] be(came) satisfied,” which he believes was pronounced *raṣ₂iya (like Classical Arabic raḍiya رَضِيَ ). If Ṣafaitic *raṣ₂iya did not end with –a, then it would have sounded *raṣ₂iy, and the terminal *–iy = *–ȋ would not have been written. However, is there any evidence that Ṣafaitic rṣ₂y had the vowel *i before the third radical? If I were to guess, I would say that it had the vowel a rather than i (I will mention the reason later when I talk about verbs III=w/y). Currently I do not know enough to express an opinion on the terminal short vowels of Ṣafaitic, but the claim by Lipiński (ibid. p. 167) that Greek spellings like Μασαχηλος (Masachēlos) and Σαμαχηλος (Samachēlos) prove that Ṣafaitic perfective verbs lacked terminal –a is unconvincing. It appears that Lipiński believes that the word ʔl “lord, god” was pronounced in Ṣafaitic *ʔēl (like in Tiberian Hebrew), but this is impossible. The Ṣafaitic word must have been pronounced with a short vowel *ʔil (I will talk later about the etymology of this word). If this was preceded by a vowel a, the whole sequence *aʔil could have been rendered by Greek ηλ (ēl). So Lipiński’s argument is invalid.
§II. Given that terminal –a does not usually fall in Classical Arabic, I am assuming that terminal *–a.wa and *–a.ya changed first to *–o.wo and *–e.ye and then to *–ō and *–ē. This is analogous to the development described in §II.21.A.h.δ.. Thus, *na.ma.wa > *na.mo.wo > *na.mō > *na.mo > na.ma نَمَاْ “[he] grew (intr.),” *ba.na.ya > *ba.ne.ye > *ba.nē > *ba.ne > ba.na بَنَىْ “[he] built,” *yaħ.ya.wa > *yaħ.yo.wo > *yaħ.yō > *yaħ.yo > yaħ.ya يَحْيَاْ “[he] live(subjunctive.),” *yan.sa.ya > *yan.se.ye > *yan.sē > *yan.se > yan.sa يَنسَىْ “[he] forget(subjunctive.),” *ʔ·al-ʕa.ṣa.wa > *ʔ·al-ʕa.ṣo.wo > *ʔ·al-ʕa.ṣō > *ʔ·al-ʕa.ṣo > ʔ·al-ʕa.ṣa الْعَصَاْ “the stick, staff(acc.),” and *ʔ·al-fa.ta.ya > *ʔ·al-fa.te.ye > *ʔ·al-fa.tē > *ʔ·al-fa.te > ʔ·al-fa.ta الْفَتَىْ “the young man(acc.).”
§III. Terminal *–a.wu and *–a.yi perhaps contracted to *–aw and *–ay, and next they were affected by §II.21.A.h.η.. For example, *yaħ.ya.wu > *yaħ.yaw > *yaħ.yow > *yaħ.yo > yaħ.ya يَحْيَاْ “[he] lives,” *ʔ·al-ʕa.ṣa.wu > *ʔ·al-ʕa.ṣaw > *ʔ·al-ʕa.ṣow > *ʔ·al-ʕa.ṣo > ʔ·al-ʕa.ṣa الْعَصَاْ “the stick, staff(nom.),” and *ʔ·al-fa.ta.yi > *ʔ·al-fa.tay > *ʔ·al-fa.tey > *ʔ·al-fa.te > ʔ·al-fa.ta الْفَتَىْ “the young man(gen.).”
§IV. It is theoretically possible that terminal *–a.wi and *–a.yu contracted to *–aɥ and then *–ay, where ɥ means a labiopalatal approximant, that is, a mixed sound of w and y. According to Sibawayh (chapter 499), some of the Ṭayyiʔ طَيِّئٌ (a tribe of northwestern Naǵd) said ʔaf.ʕaw أَفْعَوْ for standard ʔaf.ʕa·y أَفْعَىْ “snake.” It is theoretically possible that the actual pronunciation of ʔaf.ʕaw أَفْعَوْ was *ʔaf.ʕaɥ. This would have been an older version of *ʔaf.ʕay which produced standard ʔaf.ʕa·y أَفْعَىْ . An alternative theoretical possiblity is that the form *ʔaf.ʕaɥ developed in some dialects to ʔaf.ʕaw rather than ʔaf.ʕay. The point against these hypotheses is that the form ʔaf.ʕaw أَفْعَوْ was not said to be a speicific nominative form. If this form was originally a nominative *ʔaf.ʕa.yu, one would expect other forms descended from the accusative *ʔa.fʕ.aya and genitive *ʔa.fʕ.ayi.
According to a source quoted in the Lisȃn (e.g. under fʕȃ), the words ʔaf.ʕaw أَفْعَوْ and ʔaf.ʕay أَفْعَيْ were pronounced by some speakers ʔaf.ʕaww أَفْعَوّ and ʔaf.ʕayy أَفْعَيّ . Perhaps the real or original pronunciation was *ʔaf.ʕa.wu أَفْعَوُ and ʔaf.ʕa.yi أَفْعَيِ . The dialects which had these forms would have been dialects in which the terminal vowels –u and –i were not lost, so that *–a.wi and *–a.yu developed to *–a.ɥi or *–a.ɥu rather than *–aɥ.
§V. In standard CA the etemological endings *–a.wV# and *–a.yV# still contract to –ȃ when pronominal suffixes are attached to them, e.g. *qa.fa.yu-ka > qa.fȃ-ka قَفَاْكَ “your(masc. sing.) nape” and *ʔarayu-ka > ʔarȃ-ka أَرَاْكَ “[I] see you(masc. sing.).” (Note, however, that etemological *–aw# and *–ayV# retain their original form with pronominal suffixes, e.g. namaw-tu نَمَوْتُ “[I] grew,” bakay-tu بَكَيْتُ “[I] wept,” ʔilay-ka إِلَيْكَ “to you(masc. sing.),” and ʕalay-ka عَلَيْكَ “on you(masc. sing.).“)
In some Classical dialects of the Ħiǵȃz the etemological endings *–a.wV# and *–a.yV# had the forms –aw and –ay with pronominal suffixes, e.g. qa.fay-ka قَفَيْكَ “your(masc. sing.) nape,” etc. (see Chaim Rabin, p. 87, §t). There is at least one old author who claimed that this was also the usage of the Meccan dialect (ibid.), but the orthography of the Qurʔȃn speaks otherwise. The word ʔatȃ-ka أَتَاْكَ “[he] came to you(masc. sing.)” is written in the Qurʔȃn اتيك , which can mean either ʔatey-ka (ʔatay-ka) or ʔatȇ-ka; the word banȃ-hȃ بَنَاْهَاْ “[he] built her/it” is written بنيها , which can mean either baney-hȃ (banay-hȃ) or banȇ-hȃ; the word zakkȃ-hȃ زَكَّاْهَاْ “[he] purified her/it” is written زكيها , which can mean zakkey–hȃ or zakkȇ–hȃ; the word naǵǵȃ-nȃ نَجَّاْنَاْ “[he] salvaged us” is written نجينا , which can mean naǵǵey-nȃ or naǵǵȇ-nȃ, the word ʔa.rȃ-ka أَرَاْكَ “[I] see you(masc. sing.),” is written أريك , which can mean ʔarey-ka or ʔarȇ-ka; the word yaɣ.šȃ-hum يَغْشَاْهُمْ “[he/it] covers them(masc. plur.),” is written يغشيهم , which can mean yaɣ.šey-hum or yaɣ.šȇ-hum; etc. However, the writing of words from roots III=w is unambiguous, the word daʕȃ-kum دَعَاْكُمْ “[he] called on you(masc. plur.),” is written دعاكم . This Qurʔȃnic spelling cannot mean daʕow-kum (daʕaw-kum); it probably means daʕȏ-kum. Similarly, the word ʕaṣȃ-hu عَصَاْهُ “his stick, staff” is written عصاه , which probably means ʕaṣȏ-hu, etc.
So if the Qurʔȃnic orthography represents the Meccan dialect, it seems that in this dialect the etemological endings *–a.wV# and *–a.yV# contracted to –ō and –ē before suffixes. This is basically the same as the standard language of the grammarians.
κ $Caw# > $Cay#; $Ca.wa# > $Ca.ya#
§I. An interesting development that appears to have happened in Arabic is the categorical change of terminal –aw# > –ay# and –awa# > –aya# in almost every word with more than three letters, that is to say, in almost every word longer than C₁aC₂awa. This change did not happen in terminal syllables whose first sound is y, i.e. terminal syllables of the form yaw# & yawa# were not changed to yay# & yaya#.
This shift explains why the letter ى is used in Arabic orthography to denote the etemological endings *–aw(V)# and *–ay(V)# in almost all words with more than three letters, except when the endings are preceded by y, as in *dun.ya.wu > *dun.yaw > dun.yȃ دُنْيَاْ “present world,” *ħay.ya.wa > ħay.yȃ حَيَّاْ “[he] greeted, cheered,” and *ta.ʕā.ya.wa > ta.ʕȃ.yȃ تَعَاْيَاْ “[he] feigned inability or sickness.”
———————————————————————————
۞ SIDE NOTE
The same shift seems to be found in the Ancient North Arabian inscriptions. The word nħyy “[may we] live [prosperously]” is found in inscrptions from Dadȃn. This word perhaps should be read *naħyaya or *niħyaya. It appears that in Dadȃn the shift was extended even to final syllables of the form yawa#.
In fact, it seems that some Classial dialects generalized this shift even to triliteral words. This would explain Sibawayh’s statement (§II.21.A.h.η.II.) about the pronunciation of the “shortened ʔalif“ in some dialects, and would explain why almost every verb III=w appears to have a III=y variant.
This shift is found also in Ugaritic, Hebrew, and Aramaic, but there are some exceptions in Ugaritic (e.g. Ugaritic ảs₁lw /*ʔas₁law/ for Hebrew *ʔas₁lay > ʔɛs₁lɛh אֶשְׁלֶה “[I] relax”). This shift does not exist in Ethiopic, and it appearntly did not exist in Sabaic (e.g. Sabaic hrs₂w /*hars₂awa/ for CA *ʔarḍawa > *ʔarḍaya > ʔarḍa·y أَرْضَىْ “[he] pleased”).
Based on its distribution, it seems that this shift was an areal phenomenon of the Fertile Crescent and northern Arabia. Probably it started in the Fertile Crescent and spread down to Arabia in the Iron Age.
This shift should perhaps be compared to a much earlier shift, which is the universal change of initial #wa– > #ya– in the Northwest Semitic languages. In both cases it seems that an unstressed w was universally changed to y.
———————————————————————————
§II. Some dialects of Classical Arabic may have not participated in this shift. According to Sibawayh (chapter 499), some of the Ṭayyiʔ طَيِّئٌ said ʔaf.ʕaw أَفْعَوْ for standard *ʔaf.ʕa.wu > *ʔaf.ʕaw > *ʔaf.ʕay > ʔafʕa·y أَفْعَىْ “snake.” The form ʔaf.ʕaw أَفْعَوْ either preserves the original third radical w or is a different product of an original *ʔaf.ʕa.yu (vide supra §II.21.A.h.ι.IV.).
λ $Ciw.C > $Ciy.C; $Cuy.C > $Cuw.C; $Cuy.y > $Ciy.y
§I. In the standard language the sequences $Ciw.C & $Cuy.C are prohibited. When such sequences arose through morphological derivation they were regularly changed to $Ciy.C = $Cȋ.C & $Cuw.C = $Cȗ.C, respectively.
For example, *ʔiw.ǵā.bun > ʔiy.ǵȃ.buṋ إِيْجَاْبٌ “obliging,” *ʔ·is.tiw.rā.dun > ʔ·is.tiy.rȃ.duṋ اِسْتِيْرَاْدٌ “importing,” *ʔuy.qi.ð̣a > ʔuw.qi.ð̣a أُوْقِظَ “[he] was woken up,” *yuy.si.ru > yuw.si.ru يُوْسِرُ “[he] becomes rich,” *ʔuy.qi.nu > ʔuw.qi.nu أُوْقِنُ “[I] become sure,” and *muy.ni.ʕun > muw.ni.ʕuṋ مُوْنِعٌ “ripening, ripe.”
§II. The sequence $Cuy.y becomes $Ciy.y. For exmaple, *mar.muw.yun > *mar.muy.yun (§II.21.A.h.μ.) > mar.miy.yuṋ مَرْمِيٌّ “thrown.”
§III. In some of the modern spoken dialects the sequence iw may be found in verbs I=w, e.g. yiwṣal “[he] arrives.” Other dialects have fomrs similar to the Classical yaṣilu يَصِلُ (e.g. yiṣal in Aleppo).
μ $CVw.y > $CVy.y; $CVy.w > $CVy.y
§I. The sequences $CVw.y and $CVy.w are changed to $CVy.y. For example, *law.yun > lay.yuṋ لَيٌّ “bending,” *ʕay.wun > ʕay.yuṋ عَيٌّ “inability, sickness,” *mar.mȗ.yun = *mar.muw.yun > *mar.muy.yun > mar.miy.yuṋ مَرْمِيٌّ “thrown,” and *ṣa.bȋ.wun = *ṣa.biy.wun > ṣa.biy.yuṋ صَبِيٌّ “boy.”
§II. As per §II.21.A.h.ι.V., in standard CA the etymological endings *–a.wV# and *–a.yV# contract to –ȃ when the pronominal suffix –ya “[of] me” is attached to them e.g. *ʕa.ṣa.wu-ya > ʕa.ṣȃ-ya عَصَاْيَ “my stick, staff” and *qa.fa.yu-ya > qa.fȃ-ya قَفَاْيَ “my nape.” However, in those Classical dialects which changed *–a.wV# and *–a.yV# to –aw and –ay before pronominal suffixes, the resulting forms were *ʕa.ṣaw-ya > ʕa.ṣay-ya عَصَيَّ “my stick, staff” and qa.fay-ya قَفَيَّ “my nape” (for more examples see Chaim Rabin, p. 87, §t).
§III. In modern Aleppo Arabic the sequence $Caw.y is found in ʔ̣aw.yān < *qaw.yȃn قَوْيَاْن “be(com)ing strong(er)” (from qawiya قَوِيَ “[he] be(came) strong(er)”), and the sequence $Cay.w is found in ħay.wān “animal” (from ħa.ya.wȃ.nuṋ حَيَوَاْنٌ “living; animal”).
ν $Ci.wV(V), $Cu.wV(V), $Ci.yV(V), & $Cu.yV(V)
§I. The sequences $Ci.wi, $Ci.wī, $Ci.yi & $Ci.yī were contracted to $Cȋ. For example, *ʔ·al-ʕā.di.wi > *ʔ·al-ʕā.di.yi > ʔ·al-ʕȃ.dȋ اَلْعَاْدِيْ “the aggressor(gen.),” *ʕā.di.win > *ʕā.di.yin > **ʕā.dīn > ʕȃ.diṋ عَاْدٍ “an aggressor(gen.),” *ʕā.di.wī.na > *ʕā.di.yī.na > ʕȃ.dȋ.na عَاْدِيْنَ “aggressors(gen.),” *ʔ·aθ-θā.ni.yi > ʔ·aθ-θȃ.nȋ اَلْثَّاْنِيْ “the second(gen.),” *θā.ni.yin > **θā.nīn > θȃ.niṋ ثَاْنٍ “a second(gen.),” and *tar.mi.yī > tar.mȋ تَرْمِيْ “[you(fem. sing.)] throw(irrealis).”
§II. The sequences $Ci.wu, $Ci.yu, $Cu.wi, $Cu.yi, $Cu.wī, $Cu.yī, and $Cu.yu were probably contracted to $Cü̱. In standard Classical Arabic $Cü̱ became $Cȋ. The vowel ü̱ (IPA /yː/, in Arab linguistics called إِشْمَاْمُ اَلْكَسْرِ اَلْضَّمَّ) existed in some dialects of Classical Arabic where the standard dialect has ȋ.
Examples are *ʔ·al-ʕā.di.wu > *ʔ·al-ʕā.dü̱ > ʔ·al-ʕȃ.dȋ اَلْعَاْدِيْ “the aggressor(nom.),” *ʕā.di.wun > **ʕā.dü̱n > **ʕā.dīn > ʕȃ.diṋ عَاْدٍ “an aggressor(nom.),” *ʔ·aθ-θā.ni.yu > *ʔ·aθ-θā.nü̱ > ʔ·aθ-θȃ.nȋ اَلْثَّاْنِيْ “the second(nom.),” *θā.ni.yun > **θā.nü̱n > **θā.nīn > θȃ.niṋ ثَاْنٍ “a second(nom.),” *yar.mi.yu > *yar.mü̱ > yar.mȋ يَرْمِيْ “[he] throws,” *qu.wi.la > qü̱.la > qȋ.la قِيْلَ “[he/it] was said,” *bu.yi.ʕa > bü̱.ʕa > bȋ.ʕa بِيْعَ “[he/it] was sold,” (the forms qü̱.la and bü̱.ʕa are attested in CA), tad.ʕu.wȋ > tad.ʕü̱ > tad.ʕȋ تَدْعِيْ “[you(fem. sing.)] call(irrealis)” (the forms tadʕuwȋna and tadʕü̱na are both attested in CA, see the Lisȃn under دعا), *ʔ·at-ta.bā.ku.yu(nom.)/*ʔ·at-ta.bā.ku.yi(gen.) > *ʔ·at-ta.bā.kü̱(nom./gen.) > ʔ·at-ta.bȃ.kȋ اَلْتَّبَاْكِيْ “the feigning of crying(nom./gen.),” and *ta.bā.ku.yun(nom.)/*ta.bā.ku.yin(gen.) > **ta.bā.kü̱n(nom./gen.) > **ta.bā.kīn > ta.bȃ.kiṋ تَبَاْكٍ “feigning of crying(nom./gen.).”
§III. The sequences $Ci.wū, $Ci.yū, and $Cu.wu were contracted to $Cȗ. For example, *ʕā.di.wū.na > ʕȃ.dȗ.na عَاْدُوْنَ “aggressors(nom.),” *na.si.yū > na.sȗ نَسُوْا “[they(masc. plur.)] forgot,” and *yad.ʕu.wu > yad.ʕu.wȗ يَدْعُوْ “[he] calls.”
§IV. In the word-initial position, the sequence #Cu.wȗ usually became #Cu.ʔȗ in standard Classical Arabic, and the sequence #Cu.yȗ is preserved. In other word positions both sequences were contracted to $Cȗ. Examples are ħu.wȗ.lun حُوُوْلٌ > ħu.ʔȗ.luṋ حُؤُوْلٌ “yeras,” bu.yȗ.tuṋ بُيُوْتٌ “houses,” and *yad.ʕu.wū > yad.ʕȗ يَدْعُوْا “[they(masc. plur.)] call(irrealis).”
§V. The sequences $Ci.wa(a) and $Cu.ya(a) were preserved in the word-initial position, but in other word positions they were chnaged to $Ci.ya(a). For example, ʕi.wa.ḍuṋ عِوَضٌ “substitute; compensation,” ħi.wȃ.ruṋ حِوَاْرٌ “dialogue, conversation,” qu-yȃ.ʔuṋ قُيَاْءٌ “much vomiting,” *ʔ·al-ʕā.di.wa > ʔ·al-ʕȃ.di.ya اَلْعَاْدِيَ “the aggressor(acc.),” *ʕā.di.wan > ʕȃ.di.yaṋ عَاْدِيًاْ “an aggressor(acc.),” *ʕā.di.wā.ni > ʕȃ.di.yȃ.ni عَاْدِيَاْنِ “two aggressors,” *ʔ·at-ta.bā.ku.ya > ʔ·at-ta.bȃ.ki.ya اَلْتَّبَاْكِيَ “the feigning of crying(acc.),” and *ta.bā.ku.yan > ta.bȃ.ki.yaṋ تَبَاْكِيًاْ “feigning of crying(acc.).”
§VI. The sequences Cu.wa(a) and Ci.ya(a) are stable. For example, yad.ʕu.wa يَدْعُوَ “[he] call(subjunctive),” yad.ʕu.wȃ.ni يَدْعُوَاْنِ “[they (dual)] call,” ʔ·aθ-θȃ.ni.ya اَلْثَّاْنِيَ “the second(acc.),” θȃ.ni.yaṋ ثَاْنِيًا “a second(acc.),” yar.mi.ya يَرْمِيَ “[he] throw(subjunctive),” and yar.mi.yȃ.ni يَرْمِيَاْنِ “[they (dual)] throw.”
ξ C.wV(V)$ & C.yV(V)$
The old linguists (as well as modern ones) described phonetic changes involving wV(V) and yV(V) when preceded by closed syllables. However, the stated rules look to me dubious.
§I. One supposed rule is that C.wa and C.ya became Cȃ in words such as, **yax.wa.fu > ya.xȃ.fu يَخَاْفُ “[he] fears,” **yah.ya.bu > ya.hȃ.bu يَهَاْبُ “[he] fears,” **ʔax.wa.fa > ʔa.xȃ.fa أَخَاْفَ “[he] scared,” **maq.wa.mun > ma.qȃ.muṋ مَقَاْمٌ “place (of standing),” **mas.ya.run > ma.sȃ.ruṋ مَسَاْرٌ “path,” and **si.taq.wa.ma > ʔ·is.ta.qȃ.ma اِسْتَقَاْمَ “[he] was/became straight.”
However, there are too many exceptions to this phonetic rule, e.g. ʕad.waṋ عَدْوًاْ , qab.waṋ قَبْوًاْ , naħ.waṋ نَحْوًاْ , raʔ.yaṋ رَأْيًاْ , saʕ.yaṋ سَعْيًاْ , maš.yaṋ مَشْيًاْ , nax.wa.·tuṋ نَخْوَةٌ , naʕ.wa.·tuṋ نَعْوَةٌ , nad.wa.·tuṋ نَدْوَةٌ , daʕ.wa.·tuṋ دَعْوَةٌ , qar.ya.·tuṋ قَرْيَةٌ , ruq.ya.·tuṋ رُقْيَةٌ , ruʔ.ya.·tuṋ رُؤْيَةٌ , fir.ya.·tuṋ فِرْيَةٌ , miq.wa.luṋ مِقْوَلٌ , miṭ.wa.luṋ مِطْوَلٌ , miʕ.wa.luṋ مِعْوَلٌ , maṣ.ya.fu مَصْيَفٌ , Mid.ya.nu مِدْيَنُ , ʔas.wa.du أَسْوَدُ , ʔaħ.wa.lu أَحْوَلُ , ʔah.ya.fu أَهْيَفُ , ʔaɣ.ya.du أَغْيَدُ , etc.
§II. Other supposed sound changes are C.wu > Cȗ and C.yi > Cȋ in words such as **yaq.wu.lu > ya.qȗ.lu يَقُوْلُ “[he] says,” **yas.yi.ru > yasȋru يَسِيْرُ “[he] walks,” **mas.yi.run > ma.sȋ.ruṋ مَسِيْرٌ “waking,” etc. Also C.wū > Cȗ and C.yū > Cȋ in **max.wū.fun > *ma.xȗ.fuṋ مَخُوْفٌ “feared” and **mab.yū.ʕun > ma.bȋ.ʕuṋ مَبِيْعٌ “sold,” etc.
§III. The sequences C.wȋ, C.yȋ, C.wȃ, and C.yȃ are stable, e.g. taṭ.wȋ.luṋ تَطْوِيْلٌ “elongation (tr.),” tas.yȋ.ruṋ تَسْيِيْرٌ “making walk,” taǵ.wȃ.luṋ تَجْوَاْلٌ “wandering,” tas.yȃ.ruṋ تَسْيَاْرٌ “wandering,” etc.
§IV. It is possible that verbs such as yaxȃfu يَخَاْفُ , yaqȗmu يَقُوْمُ , and yasȋru يَسِيْرُ never had a third radical in the middle, that is, they are biradical verbs that never became triradical. Derived forms such as ʔaqȃma أَقَاْمَ , ʔ·istaqȃma اِسْتَقَاْمَ , and maqȃmuṋ مَقَاْمٌ seem to have been built by adding affixes to a basic biconsonantal stem –C₁VVC₂–. In my opinion, verbal forms based on roots II=III such as ʔamarra أَمَرَّ , ʔ·istamarra اِسْتَمَرَّ , and mamarruṋ مَمَرٌّ are also directly derived from short –C₁VC₂C₂–. The variation in the vowel quality between yaqȗmu يَقُوْمُ and maqȃmuṋ مَقَاْمٌ etc. is perhaps of morphological derivational origin. It it not impossible that the ancient Semitic-speakers changed the quality of the long vowel in the verbal stem –C₁VVC₂– in accordance with morphological rules.
ο w/y > ʔ
§I. The development wi– > ʔi– is seen in the words ʔirθuṋ إِرْثٌ and ʔirȃθa·tuṋ إِرَاْثَةٌ , which are variants of wirθuṋ وِرْثٌ and wirȃθa·tuṋ وِرَاْثَةٌ , all of which mean “inheritance” (from the roo wrθ). Other examples are ʔiǵȃduṋ إِجْدَاْنٌ for wiǵȃduṋ وِجْدَاْنٌ “emotion,” ʔiʕȃʔuṋ إِعَاْءٌ for wiʕȃʔuṋ وِعَاْءٌ “vessel,” ʔišȃħuṋ إِشَاْحٌ for wišȃħuṋ وِشَاْحٌ “scarf,” ʔikȃfuṋ إِكَاْفٌ for wikȃfuṋ وِكَاْفٌ “saddle (of animals other than the horse),” and ʔisȃda·tuṋ إِسَاْدَةٌ for wisȃda·tuṋ وِسَاْدَةٌ “pillow.”
The development wu– > ʔu– is seen in ʔuʕida أُعِدَ for wuʕida وُعِدَ “be(came) promised,” ʔuħiya أُحِيَ for wuħiya وُحِيَ “be(came) inspired, revealed,” ʔuqqita أُقِّتَ for wuqqita وُقِّتَ “be(came) timed, fixed, determined,” ʔukȃfuṋ أُكَاْفٌ for wukȃfuṋ وُكَاْفٌ “saddle (of animals other than the horse),” and ʔuǵȗduṋ أُجُوْدٌ for wuǵȗduṋ وُجُوْدٌ “existence.”
The author اَلْزَّمَخْشَرِيُّ claimed that the change of initial wu– to ʔu– was possible in any word, and اَلْمَاْزِنِيُّ claimed the same thing for the change of initial wi– to ʔi–. Based on these claims it would seem that these changes were in reality an elision of the word-initial semivowel w. This was typical of the Akkadian language, where all word intitial w and y were dropped (thus Old Akkadian yiqīs “[he] presented” and yuballiṭ “[he] kept alive” correspond to Babylonian and Assyrian iqīs and uballiṭ. Old Akkadian waθābum “sitting” corresponds to Middle Babylonian ašābu and to Middle Assyrian usābu).
———————————————————————————
۞ SIDE NOTE
According to Antonio Loprieno (in Ancient Egyptian: A linguistic introduction (1995), pp. 34-35), Ancient Egyptian ı͗ 
was pronounced y (palatal approximant) in the Old Kingdom, but in the Middle Kingdom this became ʔ when it was initial and unstressed. For example, ı͗wn “color” was pronounced *ya.ˈwin in the Old Kingdom and *ʔa.ˈwin in the Middle Kingdom.
Since this was a regular rule, it is possible that an initial unstressed y was just dropped, so that the supposed *ʔa.ˈwin was really pronounced *a.ˈwin.
———————————————————————————
§II. The word taʔrȋxuṋ تَأْرِيْخٌ “dating, history” has a variant tawrȋxuṋ تَوْرِيْخٌ . The original root of these words is wrx, as can be seen from Akkadian warxum “moon; month,” Ethiopic warx “moon,” Ugaritic yrx < *wrx “moon,” Hebrew yārēaħ יָרֵחַ < *warix– וַרִח* “moon,” and Syriac yarħâ ܝܰܪܚܳܐ < *warix– ܘܰܪܚ* “month.” It would seem that taʔrȋxuṋ (which is a Form II verbal noun) is derived from the Form II verb ʔarraxa أَرَّخَ “dated, chronicled,” while tawrȋxuṋ is derived from the variant warraxa وَرَّخَ “dated, chronicled.” Thus, the original variation could have been between ʔarraxa and warraxa. Since the form ʔarraxa is more common than warraxa, it is possible that the second form is a back-derivation from a western Arabian imperfective yuʔarrixu > yuarrixu > yuwarrixu يُوَرِّخُ (by elision of the hamza·t and addition of a secdondary glide, see §II.21.A.i.η.). Since the root ʔrx is attested in Arabic only in the meaning “dating, history,” it seems to be a loan from another Semitic language. In my opinion, it is possible that this root is a loan from the Babylonian language, because the phonetic change wa– > a– happened regularly in Middle Babylonian (while it is unusual in Arabic as far as I know). The forms warxu and arxu for “moon, month” are both attested in Babylonian.
The word ḍawḍȃʔuṋ ضَوْضَاْءٌ “noise” has a variant ḍaʔḍȃʔuṋ ضَأْضَاْءٌ . It is possible that the change in this case is just a simple assimilation of the first syllable to the second one.
§III. The sequences a.wȗ and u.wȗ often become a.ʔȗ and u.ʔȗ, e.g. *qawūlun > qaʔȗluṋ قَؤُوْلٌ “much saying,” *ṣawūlun > ṣaʔȗluṋ صَؤُوْلٌ “bully,” *ħuwūlun > ħuʔȗluṋ حُؤُوْلٌ “be(com)ing a barrier,” and *ɣuwūrun > ɣuʔȗruṋ غُؤُوْرٌ “be(com)ing like a cave; going deep inside.”
It would seem that the sequence a.wȗ/u.wȗ was reduced first to a.ȗ/u.ȗ, that is, the sound w was elided and replaced by a hiatus (which Sibawayh called the “intermediate hamza·t” اَلْهَمْزَةُ اَلَّتِيْ بَيْنَ بَيْنَ , see §II.20.A.b.). Next this was fortified to a true hamza·t.
The sequence a.yȗ is stable, e.g. hayȗbuṋ هَيُوْبٌ “much fearing” and bayȗʕuṋ بَيُوْعٌ “much selling.”
§IV. A regular change of the standard langauge is ā.w > ȃ.ʔ and ā.y > ȃ.ʔ whenever the w and y were unstressed. For example, *sa.ˈmā.wun > samȃʔuṋ سَمَاْءٌ “sky, heaven,” *bi.ˈnā.yun > binȃʔuṋ بِنَاْءٌ “building,” *ḍaw.ˈḍā.wun > ḍawḍȃʔuṋ ضَوْضَاْءٌ “noise.” Compare the dual forms sa.mȃ.ˈwȃ.ni سَمَاْوَاْنِ “two heavens” and bi.nȃ.ˈyȃ.ni بِنَاْيَاْنِ “two buildings.”
Again, this change could have been originally ā.wV/ā.yV > ā.V. The hiatus was then filled by a real hamza·t.
It seems that the change ā.w/ā.y > ȃ.ʔ was characteristic of western Arabia (see Chaim Rabin, p. 143, §kk). This would make sense, because the western Arabian dialects are known for having lost the hamza·t sound (vide §II.21.A.i.γ.). If they elided the hamza·t, they could have also elided the semivowels w and y. A question arises whether the ʔ < w/y ever existed in real speech. The western Arabian dialects only had the “intermediate hamza·t” اَلْهَمْزَةُ اَلَّتِيْ بَيْنَ بَيْنَ . In reality, words such as سَمَاْءٌ and بِنَاْءٌ could have been pronounced samȃuṋ and binȃuṋ, i.e. without a real hamza·t.
§V. In the standard language of the grammarains the rule of the change āw/āy > ȃʔ is strictly applied except when the sounds āw and āy are followed by the suffix –a·t. Words that end with –ȃwa·t/–ȃya·t are common in the standard language, e.g. sa.ˈmȃ.wa.·tuṋ سَمَاْوَةٌ “roof, rising,” si.ˈqȃ.ya.·tuṋ سِقَاْيَةٌ “giving of water,” ʕa.ˈbȃ.ya.·tuṋ عَبَاْيَةٌ “cloak,” ʕa.ˈð̣ȃ.ya.·tuṋ عَظَاْيَةٌ “a kind of lizard,” etc. Sometimes there are variants, e.g. the mentioned examples have the variants samȃʔa·tuṋ سَمَاْءَةٌ “roof, rising,” siqȃʔa·tuṋ سِقَاْءَةٌ “giving of water,” ʕabȃʔa.·tuṋ عَبَاْءَةٌ “cloak,” and ʕað̣ȃʔa·tuṋ عَظَاْءَةٌ “a kind of lizard.” The linguist اِبْنُ السِّكِّيْتِ wrote that the variants with –ȃʔa·t belonged to the speech of Ħiǵȃz (see Chaim Rabin, p. 143, §kk). This is a very useful information. It is surprsing that this is (apparently) the only author who mentioned it. Since we know that the speech of Ħiǵȃz lacked the hamza·t sound, we should perhaps assume that the written –ȃʔa·t was in reality pronounced –ˈȃ·t with a two-peak stress.
الراصد العربي 