Vocalisation in Group Writing: A New Proposal

The so-called group writing or syllabic orthography is a special orthography used in Egyptian hieroglyphic texts starting from the New Kingdom/Late Bronze Age. The nature and function of this orthography, especially the way it notates vowels, has been a topic of debate for more than a century, without any consensus being reached. In this book, Marwan Kilani presents a new interpretative model that provides a fresh explanation of how the syllabic orthography notates vowels. The author starts from a critical reanalysis of previous suggestions and from a thorough reassessment of the evidence. He then infers the functioning of the system by comparing the group writing spelling of Late Egyptian words surviving in Coptic with the reconstructions of their vocalizations. This approach leads to the recognition of a system that not only coherently explains all the spellings attested in the corpus, but which also produces interpretations of the spellings in group writing that agree with current reconstructions of the Egyptian vocalization. The book contains indexes and an Appendix listing the words analysed in the study.

It is easier to understand the interpretative model presented in this book if one knows how the whole system looks like and works since the beginning.
In this foreword, therefore, I provide a brief but exhaustive description of the functioning of group writing according to my new proposal. The reader who is simply interested in knowing how to interpret group writing words can just read this foreword, and ignore the rest of this study.
As for those who want to understand why my system looks like this, what its theoretical bases are, how the evidence supports it, and how it can be used to confi rm the current reconstructions of the Egyptian vocalisation, they will fi nd all this information discussed in detail in the following chapters.
I divide this forward into 10 points, each illustrating a specifi c feature of my system. 1 • In general, words written in group writing are transcribed according to their contemporary vocalisation. This means that the specifi c phase of the language must be taken into account when extrapolating the vocalisation, and it also means that a same word can be spelled diff erently in diff erent periods, if some relevant phonological change took place between such periods.
2 • The group writing orthography indicates only two vowels, or better two vocalic classes, namely a back vowel class, transcribed in the present book as U, and a non-back vowel class, transcribed here as A.
The unstressed vowels are often diffi cult to reconstruct, and therefore diffi cult to analyse. For this reason this book will focus exclusively on stressed vowels. As a preliminary observation, however, it is worth mentioning that the available evidence suggests that in earlier texts the Egyptians did perceive a distinction between non-back and back vowels also in unstressed syllables. Moreover, it seems that the unstressed vowel /ə/ was perceived as belonging to the non-back class. Both these observations, however, would need further study to be confi rmed. 4.a • Vowels belonging to the back vowel class are always indicated by the presence of a w, which is usually written either with the sign = w or with a biliteral sign/group whose second consonant is w, such as . When the back vowel is associated with the consonant /t/, the spelling is also possible (see Kilani 2017a, 200-1). When the back vowel is associated with a consonant /k/, the group could also be used (see below Point 10 and §4.5.1). 4.b • In group writing, vowels belonging to the non-back vowel class are indicated by the presence of a Ꜣ, which is usually written either with the sign = Ꜣ or with a biliteral sign/ group whose second consonant is Ꜣ, such as . In addition, the absence of any marker (either Ꜣ or w) and the use of single consonantal signs can also indicate a consonant associated with a non-back vowel. So for instance, the sequences *sa or *si would both be interpreted as sA and in principle could be written as , or . When the non-back vowel is associated with the consonant /t/, the group is usually used. 4.c • The marker Ꜣ is also used in cases characterised by the absence of any vowel. 1 In these cases, I transcribe the Ꜣ as 0 (= zero), rather than as A. 5 • In contrast with what usually assumed in previous scholarship, the sign does not indicate any vowel. Rather, it is a sort of diacritic sign used to modify the pronunciation of the consonant or group to which it is associated. In order to highlight this function as diacritic I transliterate it as ʸ. The nature of such modifi cations seems to depend on the sign or group to which is associated (see below §4.2). For instance, when associated with = ḫ in the form = ḫʸ, it indicates a pronunciation ḫ₂ > Sahidic ϣ (exclusively), in opposition to the standard pronunciation ḫ₁ > Sahidic ϩ and ϣ. When associated with in the form = tʸ, instead, it indicates the actual presence of the phoneme /t/, i.e. it indicates that the sign has a full consonantal value and it is not just an unpronounced orthographic element. When associated with the group in the form = ʸr0, it indicates a pre-consonantal /r/, i.e. it indicates that no vowel follows the /r/, as already noticed by previous scholars. A few other cases, however, are still unclear, and would need further specifi c investigation (see below). 6 • Groups marked with c+w (where c = any consonant), namely groups characterised by a consonant + a back vowel, can be read both as cU and as Uc. A group like can thus correspond to both rU and Ur.
By contrast, the word , "chariot", can be analysed as mA.ʸr0.k0.Ub.t(A) = mArkUbt(A), with = Ub, corresponding to the contemporary (post-20th dyn.) pronunciation *mvrkˈobt(v) > Coptic ⲃⲣϭⲟⲟⲩⲧ (with ⲟⲩ = /w/ < /b/). 1 This, in fact, suggests that the non-back vowel class indicated by the marker Ꜣ would perhaps be better understood as an absence-of-back-vowel class, which thus corresponds to all non-back vowels and to absence of any vowel -as in principle, the absence of a vowel is neither a back nor a non-back vowel. This distinction, however, is conceptual rather than practical, and therefore will not be discussed further in this study. It may be, however, an interesting topic for further research, as it provides a clue about how the Ancient Egyptians perceived their own vocalic system, and their language as a whole. This is clearly the most innovative aspect of my proposal. The idea that a same syllabic group may encode both a sequence cU and Uc may at fi rst appear surprising and even counterintuitive, but as I explain in detail here below, similar pronunciation pairs for the same sign(s) are sporadically attested in other writing systems and in Egyptian itself. The Egyptian evidence provides also clues about the possible origins of such phenomenon. 7 • A sequence with a back vowel located between two identical consonants is usually transcribed with the reduplication of the same group, which has then to be read as cU.Uc. For instance, the word , "oven", can be analysed as tA.rU.Ur = tArUr, with = rU.Ur = rUr. Such a spelling corresponds to a contemporary pronunciation *tvrˈu:r(v), deriving from Sem. *tv(n)nu:r(v) 2 and developing into Coptic ⲧⲣⲓⲣ (with regular /u:/ > /i:/ = ⲓ in proximity of /r/). 8 • As already observed by previous scholars, the groups and transcribe exclusively word-fi nal or pre-consonantal /n/ and /r/, i.e. /n/ and /r/ not followed by any vowel. In addition, the group can be combined with a following /r/ to transcribe /l/. However, as I discuss below ( §4.3) and in contrast with what assumed by previous scholars, I believe that such groups should not be understood as codas of the previous syllable. Rather, from the point of view of the graphic (not linguistic) syllabifi cation, it looks like the Egyptians perceived them as connected with the following group, clustered with its initial consonant. Therefore, for instance, the above-mentioned word was likely perceived and syllabised by the Egyptians as sA-rpU-tA, rather than as sAr-pU-tA, as most speakers of European languages would tend to do.
Although apparently trivial, this distinction becomes crucial in light of point 6 above: since groups characterised by a back vowel can be read both as cU and as Uc (where again c = any consonant), and since these preconsontal n and r clustered with the initial consonant of the following group, then it can be inferred that groups characterised by a back vowel and preceded by or could be read as ncU / rcU or as Unc / Urc , but not as *nUc / *rUc, as one would expect if and were perceived as codas of the previous syllable.
Therefore, a word like 4 Foreword 10 • Egyptian vowels /a/ and /a:/ after /k/ appear to have shifted to /o/ and /o:/, or at least to have been realised and perceived as back vowels, already during the 18th Dynasty, if not before (see below §4.5.1). This means that after a /k/, a back vowel must be expected even in periods before the general /a:/ > /o:/ and /a/ > /o/ shifts took places. Note that instances of such early shift had already been occasionally noticed by various scholars. An example that is often mentioned is the word , which is transcribed as ku in Cuneiform texts even in periods when we would expect a pronunciation, and a Cuneiform transcription, *ka.
As I will discuss in this book, the points just described are all derived from observations based on the Egyptian or Coptic evidence, and the resulting system is fully coherent and can be applied throughout the whole of the New Kingdom. At the same time, this system allows to analyse the Egyptian vocalisation through native Egyptian sources for the fi rst time, and the fact that the readings obtained through it are overall in agreement with the reconstructions advanced so far confi rms its general validity.
6 §1 Introduction logical framework and a sketch of a new model that can be the starting point for further future research. §2 Methodological problems in previous models Usually, the main issues aff ecting the solutions suggested so far are related with their methodological frames and with the a priori assumptions on which they are based. A fi rst problem aff ecting all previous interpretations is the assumption that a system recognising and encoding three vowels /a(:)/, /i(:)/, and /u(:)/ underlies the group writing orthography, as it is the case for various Semitic scripts. Such assumption, however, is not based on any concrete evidence from the contemporary Egyptian texts. Although it is true that the original Middle Egyptian vocalic system was probably phonologically based on a three-vowel opposition, there is no real evidence indicating that this system was still preserved, phonologically and/or phonetically, during the New Kingdom. In fact, the evidence that we have seems to indicated that during the New Kingdom the vocalic system of Egyptian was already moving toward the Coptic one (e.g. Loprieno 1995, 38-9). In addition, and more important, the Egyptian perception of their vocalic system could have actually been very diff erent from both its phonological and phonetic realities, and even if the vocalic system of the New Kingdom were indeed still a tri-vocalic system, there is no reason to assume that the Ancient Egyptians themselves perceived and conceptualised it as such. As consequence, there is no reason to assume a priori that the Egyptians felt the need to indicate in writing exactly these three vowels. In fact, they could have also developed a system in which more vowels were indicated 7 or one in which less than three vowels were graphically distinguished. 8 Such a discrepancy between the perception of the vocalisation and its phonological reality is not only relatively common in many written languages and writing systems around the world, 9 but it is also attested in Coptic. 10 Therefore, there is no reason to assume, a priori, that the Egyptians perceived as distinct vowels only and exactly the three vowels /a/, /i/ and /u/, and therefore there is no reason to assume, a priori, that the vocalic values of group writing must refl ect this tri-vocalic division.
Another problematic and somehow related assumption aff ecting previous interpretations of group writing is the idea that Ꜣ, w and y, either as independent signs , and or as second consonant in biliteral groups (in the case of Ꜣ, w), are all vocalic markers, or more in general that they all have to be interpreted as graphemes representing distinct phonemes or phones. Other possibilities should also be considered. For instance, one or more 8 §2 Methodological problems in previous models of these signs may have been diacritics, and they could have been used for indicating some modifi cation of the phonetic -consonantal or vocalic-value of an associated grapheme, or they could even have represented the absence of a feature or phoneme. 11 Again, there is no reason to assume a priori that these three elements must necessarily be vocalic markers.
Besides these two theoretical problems, two additional methodological issues often fl aw past interpretations of group writing. The fi rst is the fact of focusing mainly (Albright 1934) or exclusively (Hoch 1994) on words of Semitic origin. Although it is true that Semitic loanwords do represent by far the majority of words written in this orthography, they also present a series of specifi c problems that makes them generally unsuitable to identify any meaningful patterns that may underlie the system.
First of all, the vowels and vocalic structures of Semitic words are usually not fi xed, and can be changed to express diff erent grammatical forms or derived meanings. In addition, we generally do not know the exact Semitic language(s) or dialect(s) at the origin of the Semitic borrowings in Egyptian. Are these words from southern Canaanite, northern Canaanite/proto-Phoenician, Akkadian, Ugaritic, Amorrite? Or do they come from some other Semitic language or dialect poorly or not attested at all? Egypt had direct contacts with various North-West-Semitic dialects of the Levantine coast and Syria, and Egyptian scribes used Akkadian as lingua franca: Semitic borrowings could thus virtually come from any of these languages. Moreover, even if we knew the exact language or dialect from which the Egyptians took these forms, the attestations of these languages (except Akkadian and Ugaritic) and of their vocalisation during the Late Bronze age are so scanty that any meaningful comparison would be extremely diffi cult. We could obviously work with theoretical reconstructions, but in this case we have to consider that if on the one hand reconstructions can be very precise on the phonological level, on the other they do not tell us anything about the phonetic realisations of phonemes involved, i.e. about the actual pronunciation that the Egyptians would have heard or perceived.
In most cases it is also impossible to determine when the word was borrowed, as such words may have entered Egyptian decades or even centuries before their earliest attestation in the texts. This is a crucial issue, because during that span of time phonetic changes leading to divergent vocalisations may have occurred both in Egyptian and in the donor language.
Many of the solutions advanced so far have paid little or no attention to such synchronic or diachronic considerations. Nevertheless, these are crucial methodological issues, because any approach that does not take them into account can only produce a huge corpus of words that, however, is not internally coherent and from which, therefore, no coherent result can be expected.
Some of these considerations are at the origin of Zeidler's proposal (1991) for a new approach, mentioned also by Peust in his review of previous scholarship on the topic (1999, 221). In particular, Zeidler recognises the problem of using Semitic forms, and decides to focus instead on those words written in group writing which are attested also 9 §2 Methodological problems in previous models in Coptic. In other words, he proposes a change of perspective and he suggests to analyse the words written in group writing starting from their Coptic descendants, rather than from their supposed Semitic prototypes (see also Peust 1999, 221). As he says, using Coptic allows analysing the group writing from an "inner-Egyptian" perspective, bypassing all the incertitude deriving from the borrowing process highlighted above.
I think this is an excellent solution, from a methodological point of view. Zeidler application of this idea, however, presents problems that need to be addressed. First, on the basis of his article, it seems he did not distinguish the Egyptian words on the basis of the periods in which they are attested. This is a serious issue, which essentially fl aws the whole study: group writing is attested over a long period of time, during which various phonological developments took place in Egyptian, especially at the vocalic level. To ignore these phonological changes and to compare all the words only with their Coptic descendants is therefore methodologically problematic: on the one hand it is clear that many of the Egyptian forms may refl ect vocalisations that are diff erent from those attested in Coptic, while on the other hand, since these Egyptian forms may come from diff erent periods, that is from diff erent "phonological phases", the way their vocalisation relates with the corresponding Coptic forms may change from one word to the other, if their attestations are not synchronic. For instance, we know that the phonological change /a:/ > /o:/ likely took place just after the reign of Ramses II (Loprieno 1995, 38). It is therefore to be expected that in words from before Ramses II, Coptic /o:/ will be transcribed as a non-back vowel, while in those after his reign it will correspond to a back vowel. However, if words from both periods are compared together, without chronological distinction, as Zeidler seems to do, then an inconsistency has to be expected, as it will seem that the same vowel could be transcribed in two diff erent ways without apparent reason.
In addition, Zeidler, follows the previous scholars in assuming a vocalic system based on the three vowels /a/, /i/, /u/, without considering that, as said, the Egyptians' perception and conceptualisation of their vowels could have been diff erent. Similarly, he also assumes that Ꜣ, w and y act as vocalic markers, without considering any other possible functions.
Another problem in Zeidler's work is the inclusion of both nouns and verbs in his corpus. The inclusion of nouns is not an issue: their morphological variability is relatively limited and the evolution of their vocalic patterns is relatively well understood. Verbs, however, are problematic from many points of view. First, Egyptian verbal morphology is complex and includes various forms that were likely vocalised in diff erent ways. Therefore, to be meaningful, any comparison between Egyptian verbs and their Coptic descendants should be strictly limited to corresponding morphological forms. This, however, is often impossible, because various verbal forms attested in Late Egyptian did not survive into Coptic. For this reason I think that verbs should be excluded from any preliminary work on group writing: it is clear, in fact, that the incertitude about both the vocalisation of their Late Egyptian forms and their relation with the Coptic attestations is likely to fl aw any model built on them.
Finally, Zeidler did not publish the corpus on which his analysis is based, and he only provided a few selected examples for some of the groups he studied. This is a serious 10 §2 Methodological problems in previous models shortcoming, because it makes it impossible to verify his data, and therefore the validity of his comparisons and results.
It thus appears that various problems aff ect the models presented so far, either because of the theoretical assumptions they are built on, or because of the composition of the corpora they used, or because of both. Since these issues are essentially methodological, they should be taken into consideration in any new attempt to analyse the group writing.
The interpretation I am presenting in this book does that, as I discuss here below. §3 Group writing -A new approach §3.1 Defi nition of the corpus As already suggested by Zeidler, working with words attested both in group writing and in Coptic is probably one of the best possible solutions, from a methodological point of view. Such words, therefore, are at the basis of my corpus as well. 12 A few important diff erences, however, distinguish my dataset from his. First, only forms for which the vocalisation can be reconstructed with reasonable certainty have been considered. Verbs have therefore been excluded, for the reasons discussed above. As for nouns, only those with a reasonably likely direct descendant in Coptic have been considered. For instance, Egyptian forms which are related with Coptic words, but do not seem to be their direct ancestors, have been excluded. 13 Similarly, words attested only in diff ering morphological forms (especially singular versus plural) have also been excluded, except when the corresponding forms can be reconstructed with some certainty. Moreover, and this is the most important diff erence with Zeidler's approach, the words of my corpus have been distributed into three chronological groups, corresponding to three phonological phases divided by major phonological developments of the vocalisation. Therefore, my model is built on the comparison of the attested Egyptian words with the corresponding contemporary vocalised forms as they can be reconstructed from Coptic, rather than with the Coptic forms themselves.
26 words of my corpus belong to the fi rst period. 49 instead are attested in the second, while 23 in the third. 17 words are attested in more than one period. Of these, 5 are attested in all the three periods.
The fi rst group consists of words attested before or during the reign of Ramses II, and presents a vocalic landscape which is essentially that of Middle Egyptian, except for a possible shift /i/ > /e/ (Loprieno 1995, 38). The second groups, instead, consists of words attested in the period going from the reign of Merenptah to the end of the 20th Dynasty, and it is characterised by the phonological shift /a:/ > /o:/, which likely took place at the end of the previous period, from around 1200 BCE (Loprieno 1995, 38). Finally, Period 3 is characterised by two additional phonological changes, namely the merging of /e/ and /u/ into /e/, and the shift of /a/ into /o/. It is usually assumed that the merging /u/~/e/ > /e/ took place at some point around 1000-800 BCE (Loprieno 1995, 39). As for the change /a/ > /o/, it is assumed to have taken place around or after 1000 BCE (Loprieno 1995, 39, 12 I based my corpus on the words identifi ed by Černý (1976), Vycichl (1984) and Westendorf (1965) in their etymological dictionaries, as well as on those suggested by Ward in his review (1996) of Hoch's (1994) and Schneider's (1992) works. 13 Such as, for instance, = mrynt = "a vessel", which Černý suggests may be related with Boharic ⲙⲉⲣⲁⲛ = "tank". The connection is very doubtful, as pointed out by Hoch (1994, 137, n44), but even if the two words were really related, then the Coptic form would clearly derive from a variant that was diff erent from the form attested in group writing, because the Coptic form does not bear any trace of the y and t consonants implied by the and groups of the group writing spelling. 12 §3 Group writing -A new approach 46). In particular, this shift is already attested in the transcription of some Egyptian words in the Assyrian sources of the time of Sargon II, around 720 BCE, 14 while my study of the w-extended orthography (Kilani 2017a -see below) shows it was already in place at the time of the redaction of the tale of Wenamun, and thus possibly as early as the beginning of the 21st Dynasty.
The resulting chronological framework, and the corresponding vocalic landscape for each period, can be summarised as in tables 1-2.
Other important phonological changes aff ected the Egyptian vowel /u:/. However, both the nature of such changes and their chronological frame are rather unclear. As summarised by Peust (1999, 228-30), Coptic ⲏ seems to have transcribed two diff erent phonemes, possibly an unrounded /e:/ and a rounded /ø:/. At the same time, Coptic ⲏ appears to also derive from two diff erent phonemes, namely from an earlier /i:/ and an earlier /u:/. The relation between these two Coptic pronunciations and the two earlier vowels is not clear. It has been suggested that the fi rst are the direct refl exes of the latter, but the evidence is far from being convincing and conclusive (Peust 1999, 228-30). In addition, in many cases earlier /i:/ remains /i:/ = ⲓ until Coptic. This, actually, seems to be the regular development, while /i:/ tends to turn into /e:/ = ⲏ mainly in specifi c phonological contexts, like after /n/ = ⲛ and /m/ = ⲙ, and before /j/ = ⲓ (Peust 1999, 231-2). However, there are clear attestations of the development /i:/ > /e:/ = ⲏ also in other unexpected phononological contexts, which suggests that the picture was either less regular, or more complex (Osing 1976, I, 19-26;Peust 1999, 231-2).
It is therefore clear that these changes cannot be used to defi ne any chronological frame. The evidence emerging from the present study, however, may shed some new light on these issues of historical phonology (see below §10).   or" -C = "followed by". E.g. /e/ + -/ʔ/ = /e/ followed by /ʔ/. C-= "preceded by". E.g. /o:/ + /N/-= /o:/ preceded by /N/. C without any -= "in proximity of". E.g. /i:/ + /ʕ/~/j/ = /i:/ in proximity of /ʕ/~/j/ §3.2 The nature of the system As mentioned above, the orthographic system discussed in this book is usually defi ned by scholars as "syllabic orthography" or "group writing" (e.g. Albright 1934;Zeidler 1991;Hoch 1994;Ward 1996;Peust 1999). Both these defi nitions are descriptive, and somehow legitimate: this orthography is indeed based on a set of "groups" of signs, and it does seem 15 that each of such groups transcribed short sequences of consonants and vowels, i.e. "syllables". If however one wants to describe such orthographic system from a functional, rather than descriptive, perspective, then I think that "rebus writing" would be a more suited name. As already noticed (e.g. by Hoch 1994, 501), various "groups" are in fact nothing but short mono-or biconsonantal words (either originally native or loanword) or morphological elements and particles that happen to be used to transcribe likely homophonic syllables or segments. 16 This is evident from the fact that classifi ers may be retained when such words are used in group writing, even though they have no semantic 15 Some scholars have expressed doubts against it. Various clues, however, suggest that the system is somehow "syllabic". 16 Conceptually, such a "rebus writing" is not unique to Egyptian. A very similar principle underlies, for instance, the transcription of foreign words in Modern Chinese: "Italy", for instance, is transcribed as 意大利 = yìdàlì, which is essentially a sequence of three independent words, namely意 = yì = "meaning", "idea"; "to wish", "to desire", 大 = dà = "big", "great" and 利 = lì = "sharp", "convenient"; "profi t", which however are used purely for their phonetic value, rather than for their semantic meaning. The use of such "rebus writing" to transcribe foreign words, or even full sentences, is not a modern invention in Chinese, but is well attested also in the past. For instance, hundreds of similar transcriptions of Sanskrit terms are attested in Middle Chinese (Chen 2000), such as for instance Sanskrit maṇḍala was transcribed using the three characters/monosyllabic words 曼拏羅 , which in Middle Chinese were pronounced manH-nrae-la, and originally meant "distant", "to handle", "kind of net" respectively (Chen 2000, 395, Middle Chinese pronunciation based on and transcribed according to Baxter 2014). The most advanced development of these system, however, is probably represented by the spelling of the so-called Secret History of the Mongols, an historical chronicle about the deeds of Genghis Khan, composed in Mongolian language 15 §3.2 The nature of the system relation at all with the new word being spelled out. Let us consider, for instance, the word , "hill(s)", < Semitic √t-l-l = til(lu), 17 whose spelling can be analysed as a rebus combination of the Egyptian word t = "bread", written with its full set of classifi ers, and the Egyptian word rꜢ = "mouth". Clearly, the classifi ers of have no semantic connection whatsoever with the word "hill".
Another interesting example is the transcription of the Hurrian divine name Teshub, which appears in group writing as (Hoch 1994, 258, no. 364). This spelling can be interpreted as a rebus composed of two elements. The fi rst, t(j), possibly originates from the second person stative suffi x -t(j). The second, , is not attested as an independent word but is present in other loans from Semitic languages. Leaving aside for a moment the position of in the sequence, which will be discussed here below ( §3.3, §4.1), this can be identifi ed with a Semitic verb meaning "to return" and attested in Biblical Hebrew as šūb and in Amorite as šwb, from √š-w-b (so Hoch 1994, 258, no. 364). Once again, the is the classifi er of such verb, and it is not semantically related in any way with the name of the Hurrian god. Moreover, the presence of such classifi er shows that the sequence as to be interpreted as a single group, as in fact scholars did, rather than as a combination of two distinct groups and : does make sense as classifi er of , but it does not as classifi er of alone.
Words with a C₁-C₂ structure, 18 like , are rather rare in New Kingdom group writing. 19 At the time, the general tendency was to prefer words that in the traditional Middle Egyptian orthography would be analysed as C₁-Ꜣ (and rarely C₁-j) or C₁-w, such as sꜢ, rw, tꜢ. 20 The presence of the single stroke or of classifi ers such as in confi rms that, conceptually, the Egyptians perceived these groups as words, and not just as mere phonograms.
In Egyptian, however, the number of words with C₁-Ꜣ or C₁-w structures is limited, and they do not cover all the consonants of the language. In those cases where no suitable C-Ꜣ/w word exists, plain biliteral signs, sometimes combined with or / , or alternatively but written by using Chinese characters as purely phonetic, syllabic signs. For instance, the title of the fi rst chapter reads as follows (see Rachewiltz 2004): Chinese text : 成吉思合罕訥忽札兀兒 Pronunciation: chéng-jí-sī há-hǎ n-nè hū-zhā-wù-ér Reconstructed Middle Mongolian text: čingɣis qaɣanu hujaɣur Meaning : "(On) the origins of Gengis Khan" Conceptually, this system is very similar, if not even identical, to the Egyptian use of the group writing to transcribe sentences in foreign languages, such as in Pap. Anastasi I 23,5. 17 Akkadian tillu, Hebrew tell, Aramaic till, Syriac tell, Arabic tall -see Hoch 1994, 356-7, no. 527. 18 Where C = any consonant except Ꜣ or w. 19 While they are more frequent in earlier forms of group writing. 20 It has often been suggested that the sign was used to transcribe a third group of syllables, usually interpreted as C ₁ -y. I however disagree with this idea, as I think the evidence show that the had other functions. I discuss this aspect more in detail later, in §4.2. Therefore, for now, I focus only on C ₁ -Ꜣ (C ₁ -j) and C ₁ -w groups, which constitute the basic framework of the group orthography, as I think it will appear from this study. combination of unilateral signs if no corresponding biliteral sign was available either, could be used instead.
Single consonantal signs could also be used in the spelling of biconsonantal words (such as just mentioned), as phonetic complements for specifi c groups (such as = b-pꜢ-Ꜣ = bꜢ) or individually. When full words or bilitteral signs are available, they are clearly preferred to combinations of uniliteral signs, as it appears from Hoch index of groups (Hoch 1994, 506-12). For instance, and are by far the most common spellings of the groups ʕꜢ or rw, although in theory = ʕ + Ꜣ or r + w would have been equally valid alterntives.
Naturally, these have to be understood as overall trends, not as absolute rules. The system was clearly not fully standardised, exceptions do exist, and a certain degree of voluntary or involuntary freedom was present. Therefore, it has to be expected that at times the same consonantal sequence could be written either with a full word, with a biliteral sign, or with a combination of uniliteral signs.
Nevertheless, these theoretical considerations are crucial, because understanding how the Egyptians themselves conceived and perceived the group writing can help understand how it functioned. §3.3 The number of vowels In a previous article (Kilani 2017a) I have argued that the so-called space fi ller = w present in Late Egyptian texts was a sort of vowel marker used to give a general indication on the nature of the tonic vowel of the word. The system underlying such w-extended orthography 21 was rather rudimentary, and was characterised by two basic principles. First, the quality of only the stressed vowel was represented, and only two basic qualities, non-back and back, 22 could be indicated by the presence or absence of the respectively. Second, this vowel marker worked somehow as a classifi er, in the sense that it was written at the end of the word, after any consonant and independently from the actual position of the stressed vowel within the word. Therefore, a form like must be interpreted as consonants + vowel marker, that is as ḫpr + w => ḫwpr, corresponding to Coptic ϣⲱⲡⲉ (see Kilani 2017a, 189). The reasons for such a system, and especially for the latter feature, which may appear rather counterintuitive or at least unpractical, may have to be searched in the specifi cities of the vocalisation of Late Egyptian, and in the nature of the Egyptian writing. In particular, it is likely that Late Egyptian, like Coptic, had only one distinctive vowel, namely the tonic one, while all the other unstressed vowel were reduced and realised either as [ə] or, in some specifi c phonetic environments, as [a]. This means that only one vowel was worth being indicated, and therefore only one vocalic marker was needed for the whole word. At the same time, the fact that many hieroglyphic signs simultaneously encoded for more than one consonant, made it often practically impossible to indi-§3.3 The number of vowels cate the presence of a vowel there where it was supposed to be pronounced. For instance, as explained in Kilani 2017a, in the case of a verb like = ḫpr, "to exist", Coptic ϣⲱⲡⲉ, it would be impossible to introduce a vocalic marker w between the "ḫ" and the "p" to indicate the vowel "o" at its actual position, because the two consonants are written together with the single, indivisible triliteral sign .
If one considers that in hieratic writing some ligatures may also have been perceived as indivisible groups or "schematograms", 23 then one may understand why writing the vowel marker w after all the consonants often was not a choice, but rather the only possible option. 24 This w-extended orthography is relevant here, because it can be argued that the same principles governing it were also at the basis of group writing. In fact, if one considers that the groups of group writing were likely perceived as words within a rebus-based system, rather than just as elements of a phonetic syllabary, then one can expect them to abide to the same principles governing the spelling of ordinary words, including those underlying the w-extended orthography. These considerations constitute the basis for the interpretative model presented here. 23 Junge 2005, 38-9 -This is likely the case for words like , where the sequence of signs may have been perceived as an indivisible schematogram, thus prompting a spelling = ḫtm + w => ḫtwm, corresponding to Coptic ϣⲧⲟⲙ (see Kilani 2017a). 24 One has to remember that this system was used by people who knew the language natively. Therefore, its main aim was probably to avoid ambiguities or as an aid-memoir to facilitate the recognition of words otherwise written only consonantally, rather than to allow readers unacquainted with the language to properly vocalise new, unknown words. §4 Group writing -A new model The following sections are structured as follow: fi rst I introduce the general principles on which the proposal is based and I discuss a few special cases. The proposal will then be systematically verifi ed against the evidence from the corpus. After that, the evidence is analysed diachronically. Some words are attested in different periods, and therefore it is possible to have a closer look at the evolution of their vocalisation. The identifi cation of expected patterns of vocalic evolution can be used as a confi rmation of the validity of the model here presented.
A statistical assessment of the results is then presented, while some special cases are further discussed after that.
A general conclusion follows. §4.1 Defi nition of the interpretative model The interpretation of group writing 25 I present here is based on the two following principles, which directly derive from the w-extended orthography: -Groups encode only 2 vowels, or better 2 vocalic classes: back and non-back. 26 The presence of -w corresponds to the presence of a back vowel. In order to mark 25 In spite "rebus writing" would be conceptually more appropriate, in this study I will stick to the traditional defi nitions of syllabic orthography and group writing, just to avoid another superfl uous label to this writing system. 26 As suggested by a reviewer, since one of these vocalic classes seems to correspond to the stressed vowels /a/, /e/, /i/, while the other to the vowels /o/, /u/, in theory one may also interpret the underlying opposition as a question of roundedness/labialisation, that is as unrounded vs rounded. I think this may indeed be a valid alternative. However, I still prefer to interpret it as an opposition back vs non-back for two reasons. First, it is usually assumed that the articulatory position of a vowel is more distinctive than its roundedness (this is evident, for instance, from the fact that vocalic inventories of languages are usually described according to the position of the vowels, rather than according to their roundedness -this is also the case of descriptions of Egyptian, see for instance Loprieno 1995, passim). Moreover, there may be evidence from Coptic suggesting that the Egyptian themselves did not consider roundedness as a main distinctive feature of their vowels. In particular, as mentioned above ( §3.1), it has been suggested that Coptic ⲏ was actually used to transcribe two distinct phonemes, possibly corresponding to an unrounded /e:/ and a rounded /ø:/ (Peust 1999, 228-30). If this is the case -and I do not see any concrete reason to question it -, then the fact that the same letter ⲏ was used for both would suggest that, at least in the Coptic period, Egyptians were not distinguishing their vowels on the basis of roundedness, because roundness is actually the only main feature distinguishing the vowels /e:/ and /ø:/. Naturally, it is clear that an orthographic feature of Coptic does not tell us anything, directly, about the Late Egyptian orthography. However, I think one may expect at least some degree of continuity in the underlying perception of the speakers of the language. In particular, if Late-Egyptian/pre-Coptic speakers did perceive roundedness as a main distinctive feature of their vowels, it might be expected that, when the Greek alphabet was adopted to write Egyptian, an attempt would have been made to try to mark such a distinction, also (and perhaps especially) in the vowels /e:/ and /ø:/. This, however, is clearly not the case, which suggests that in the underlying pre-Coptic perception, roundedness was not a main distinctive feature. 20 §4 Group writing -A new model such back vowel, I use the transliteration U. The absence of -w, or the presence of -Ꜣ indicate either the presence of a non-back vowel or the absence of any vowel. I mark it with A or 0. -The -w is written after the consonant of its group, but it can be read both before or after it. This is because the -w works like a classifi er, like in the w-extended orthography. Therefore a group c-w 27 can be read either as cU or as Uc.
The possibility that only two vowels, rather than three, were encoded had already been implicitly suggested for the group writing of the Middle Kingdom (see Hoch 1994, 496-7 with refs). No one, however, seems to have considered that this principle may apply to later periods as well, possibly because of the common assumption that the sign must have also been a vocalic marker. An interpretation based on only 2 vowels, however, is strongly supported by the evidence. As it is discussed more in detail below ( §5), it can be shown that all the forms of the corpus can be explained on the basis of a two vowels system, with each group having only one vocalic value. None of the systems with three vowels suggested so far is comparably coherent.
As for the -w in group writing working like the -w in w-extended orthography, a fi rst confi rmation comes from the transcription of the name of the god Tešub mentioned above. As said, the spelling can be analysed as composed of two groups, and . The fi rst can be read as t+A = tA, where A = non-back vowel, which is the expected transcription of the fi rst syllable of the name, te-.
The second group, , instead, transcribes the syllable -šub. It is clear that if we read it as it is written, namely sbw, then the spelling cannot be reconciled with the pronunciation of the syllable it is supposed to transcribe. If instead we assume it was spelled according to the w-extended orthography, then it can be read as sb + w = sb + U, where U = back vowel, and it can be normalised as sUb, the expected transcription for the syllable -šub. A reading sUb would also fi t as a transcription of a verb /sub/ < Semitic šūb = "to return", which as mentioned above has been suggested by Hoch, on the basis of the classifi er, as the source of the group . is a C₁C₂+w group, but the same principle may be applied to C+w groups, which can thus be analysed either as C+w = C + U = CU or C+w = C + U = UC. This suggestion is supported by two words, attested in the corpus in two variants. They are the following: The spellings and can be assumed to be equivalent, as they are both attested in the same period and they are both singular, and therefore morphologically 21 §4.2 The sign identical (see Appendix A below for details). We can therefore assume that they aim at transcribing the same phonological sequence. This means that the groups and must be equivalent as well, and they also must transcribe the same phonological sequence. 28 This is possible only if we assume that the group = r+w can be read both as r+U = rU and as r+U = Ur, so that: = (r+Ꜣ) + (r+w) = (r+0) + (r+U) = r0 + Ur = rUr and: = (r+w) + (ʸr+Ꜣ) = (r+U) + (r + 0) = rU + r0 = rUr 29 The same considerations apply also to the second case: since and are morphologically the same, and come from the same period, they can be assumed to transcribe the same phonological sequence. Therefore, the groups and must be equivalent. This is possible only assuming that = d(t)+U = Ud, 30 so that: = (b+w) + (d) = (b+U) + (d+0) = bU + d0 = bUd and: These variants are relatively rare, possibly because some orthographic rules or conventions may have existed within the system (see below §7). It may appear counterintuitive that a same sign encodes for both CV and VC sequences, 31 as this is certainly not a common behaviour in attested scripts around the world. Nevertheless, a few parallels do exist. The fi rst can be found in cuneiform Hurrian, where the sign WA can be used to write both the syllables /wa/~/we/ and the syllable /aw/~/ew/. 32 In addition, a few examples may suggest a similar behaviour also for the alif+i sign in the Ugaritic alphabet, 33 but this case is more controversial (Bordreuil and Pardee 2010, 37 with refs). More important, the same phenomenon, where the reading order of the sign is inverted in respect to their writing order, is attested also in Late Egyptian, for instance in BM EA 10474 (Teachings of Amenemope), where the sequence corresponds to both wt and tw (Laisney 2007, 10).
On these theoretical bases, the system can be analysed in more details. I will do so in §5-6. First, however, a few special cases need to be introduced. 28 I.e. the same sequence of consonants and vowels, in the same order and with the stress in the same position. Obviously, it cannot be excluded that such consonants and vowels may have been phonetically realised in slightly diff erent ways, if diff erent dialects were involved, but this is not relevant here, as such variations would not aff ect the order of the phonemes themselves. 29 For the reading = ʸr+Ꜣ = r + 0 = r0 see below. 30 For = w and therefore = d(t)+U = dU/Ud see Kilani 2017a, 200-1. 31 Where V = vowel. 32 E.g. the spelling WA-ri stands for ew-ri /evri / Wegner 2007, 45. 33 Which seems to encode for both /ʔ/ + /i/ and /i/~/a/~/u/ + /ʔ/.

§4 Group writing -A new model §4.2 The sign
It has often been assumed that in group writing the sign had the function of marking a /i/ vowel. This suggestion, however, has never been demonstrated on the basis of any evidence. In fact, it has already been shown that, at least in some contexts, the sign clearly had other functions and was not used as a vowel marker. Moreover, in a system with two vowels as that implied by my model, a sign marking a third vowel /i/ would be superfl uous.
This considered, I think that the should be interpreted as a diacritic sign, which modifi es in some way the signs or groups with which it is associated. Various specifi c functions can be identifi ed from the evidence.
First, it has long been recognised by various scholars (e.g. Albright 1934, 47, 50;Helck 1971, 552-3;Schneider 1992, passim;Hoch 1994, passim) that the combination + = rꜢ, usually written in the form , is used to transcribe a word-fi nal or pre-consonantal /r/. In other instances, it appears instead that the function of is to modify the pronunciation of the consonant to which it is associated. This is the case, in particular, with ḫ = / , which corresponds to ḫ > Sahidic ϩ and ϣ, while the combination , i.e. ḫ + = ḫ₂, corresponds exclusively to ḫ₂ > Sahidic ϣ. The fact that should be interpreted as a spelling indicating a secondary pronunciation of ḫ is supported by demotic, where the sign < is specifi cally used to transcribe only ḫ₂ = Saihidic ϣ, in opposition with the sign < , which is used to transcribe a more general ḫ = Sahidic ϩ and ϣ.
The sign seems to have had also other functions. In some cases it seems it was used to indicate that the previous consonant had to be fully pronounced. This happens mainly with = r, = t, and = w, namely with consonants that at least in some phonetic or phonological environments either were not pronounced any longer, or were turned into /ʔ/. Examples from Late Egyptian can be < > = rʸrʸ = Cpt. ⲣⲓⲣ, or = ntʸ (= earlier nḏ(j) ) = ⲛⲟⲉⲓⲧ and possibly also = btʸ (= earlier bdt) = ⲃⲱⲧⲉ (see Kilani 2017a, 194), or = gw and = jswt, where the variants = gw and = jswt, spelled with , show that the of represents a fully consonantal w (Hoch 1994, 346 nos. 507, 32 27 respectively).
In the case of = dʸ, instead, the presence of seems to exceptionally indicate a group dU. This case is unique, and it is discussed more in detail here below ( §4.5.2).
The uses just described cover most of the attestations of in the corpus of this book. The role of in the remaining cases is less evident. There the function of is perhaps to provide information about the syllabic structure of the word, in order to resolve some ambiguities inherent in the system (see also §7).
To sum up, the evidence shows not only that there is no reason to assume that was a vowel marker, but also that there is no need for it to be so, as most of its occurrences can be easily explained in other ways. This indirectly supports one of the principles of the current proposal, namely that the group writing was based on a system recognising only two, rather than three, vowels. §4.3 = rC / r# and = nC / n# Many scholars (Albright 1934, 47, 50;Helck 1971, 552-3;Schneider 1992, passim;Hoch 1994, passim) have already noticed that at least two groups diff er from the others in the fact that they are functionally specialised in encoding only fi nal or pre-consonantal 24 §4 Group writing -A new model consonants. These groups are , encoding for -r.(C-) or -r# and , which I transcribe as n₁, encoding for -n.(C-) or -n#. 34 These groups appear to be diff erent from standard = r-Ꜣ and n-Ꜣ groups, in that they explicitly always correspond to r+0(+C/#) and n+0(+C/#) and never to r+A and n+A.
To mark the peculiarity of these groups, in the following paragraphs I transliterate them as ʸrꜢ and n₁ and I parse them as ʳ and ⁿ respectively.
The fact that these two consonants enjoy such a special status and treatment has parallels in other syllabic scripts. 35 This exceptionality is probably due to their phonological nature, which seems to have induced various scribal cultures to perceive them as somehow diff erent from other consonants. 36 Usually, modern scholars have assumed an implicit identity between phonological syllables and written syllables in Egyptian group writing, and therefore have interpreted the groups and as transcribing consonants in coda of (phonological) syllables, as it appears from the conventions adopted in transliterating them (Albright 1934, 47, 50;Helck 1971, 552-3;Schneider 1992, passim;Hoch 1994, passim).
These assumptions, however, have never been demonstrated on the basis of any evidence. In fact, another scenario seems to emerge from the data, which suggested that, rather than codas of previous syllables, such groups should be interpreted, from the perspective of the functioning of the script, as clustered with the following consonant.
In other words, I think the evidence shows that a sequence CV + ʸrꜢ/n₁ + CV was not interpreted by the Egyptian as CVr/n.CV, as usually modern scholars do, but was rather conceptually perceived as CV.rCV and CV.nCV. By analogy, it is likely that also when they are the last groups of a word, and must be interpreted as CV.r# and CV.n#, rather than as CVr# and CVn#.
This can be inferred from the following spelling variants attested in the corpus (II.3): All these forms can be interpreted as transcriptions of /ʕul/, from which Coptic ⲁⲗ /(ʔ)al/ derives. The spelling of Variant 1 is transparent: Where C = consonant of the following group, and # = end of word. 35 For instance, Japanese kana syllabaries have exclusively CV signs for all consonants but /n/, while in most cases Indian devanagari treats /r/ as a modifi er of other CV signs, rather than just as an independent consonant. 36 I do not know if such a phenomenon has ever been investigated from a wider ethno-anthropological perspective. Whether it has or not, I think it would be interesting to look at it integrating the Egyptian data. It is clear, however, that such a study is far beyond the scope of the present study. §4.3 = rC / r# and = nC / n# which gives ʕU.ⁿr0 = ʕU.l = ʕUl, the expected transcription for /ʕul/.
The spelling of Variant 2), needs more attention. If we assume that the signs and have to be read as codas of the previous syllables, as it has been suggested so far, then we obtain: Note that in both cases = rw cannot be read as Ur, because = -r and = -n imply a consonant just after them.
These transcriptions would thus yield, respectively: Both those readings would be problematic, both because they do not correspond in any way to the reading implied by Variant 1 and because they do not correspond to neither SBA ⲁⲗ nor F ⲉⲗ (Crum 1939, 3).
If however we assume that the signs and should be read as clustered with the following consonants, then the following interpretations become possible: Which in both cases can be read as, respectively: Which corresponds to Variant 1 and which is the expected transcription of /ʕul/.
The validity of such analysis is supported by two other words of the corpus, namely = mrḥ and = ḫ₂rd, which correspond to Coptic ⲙⲉⲣ(ⲉ)ϩ (and variants) = mˈer(ə)h and ϣⲟⲣⲧ = šˈort respectively.
= mrḥ is attested in Period 1 and 2. The words is related with Semitic rumḥ (see below §6) and its pronunciation in Egyptian can be reconstructed as *murḥv in both periods. Its spelling, therefore, can be interpreted as follow: = ḫ₂rd is instead attested in Period 3, and therefore its pronunciation at the time can be reconstructed as *ḫ₂ˈord(v). Its spelling can be parsed as: Again, taking = ʸrꜢ as the coda of the previous syllable would not work, as it would lead to a reading = mA + ʸr + ḥ-U = mAʳ + ḥU = mAʳḥU and = ḫ₂A-ʸr + d-U = ḫ₂Ar + dU = *ḫ₂ArdU, which cannot be reconciled in any way with the attested Coptic forms.
These written consonantal clusters can be phonemic, i.e. they can represent a real combination of two distinct phonemes, or can be merely orthographic, i.e. they can be a conventional combination of two graphemes used to write a diff erent phoneme, which is not independently represented within the range of available signs. This seems to be especially the case with the sign which can be combined with a following r to form a digram corresponding to Coptic ⲗ and thus probably representing /l/.
It has to be noted that the syllabifi cation dictated by these signs is a phenomenon that takes place at the writing level, and it is therefore independent from the actual syllabic and prosodic patterns characterising these words at a purely linguistic level. Similar discrepancies between the intrinsic linguistic syllabifi cation of words and the functional syllabifi cation rules commanding their written representation are attested in various writing systems around the world. English orthography, for instance, implements syllabifi cation rules that are unrelated with the actual linguistic syllabic structure of the words they 37 For = dU/Ud see below §4.5.2. 27 §4.4 in fi nal position represent, 38 while Indian devanagari (and essentially all South East Asian scripts derived from them) clusters up together all adjacent consonants irrespectively of any linguistic syllabic boundary. 39 In the latter, the resulting orthographic conventions appear to be, in this respect, conceptually and functionally very similar to what can be observed in the Egyptian group writing. §4.4 in fi nal position By the end of the Middle Kingdom the -t ending of feminine, usually written was not pronounced any longer. The use of in fi nal position, however, survives in later periods, often but not exclusively in feminine words, and it is occasionally attested also in words written in group writing. In many cases this vestigial may have been just a graphic phenomenon, possibly reinterpreted as a semantic marker of feminine grammatical gender. The evidence from the corpus used in the present study, however, seems to suggest that in other cases it may have retained some phonetic signifi cance, and may have developed into a marker for /ə/ at the end of words. Such a development would not be surprising, as the disappearance of the feminine ending -t caused the previous usually unstressed vowel to appear in fi nal position as a /ə/.
This considered, in the present study I transliterate this fi nal as t₁ and I parse it as ə, thus for instance: = šnft₁ = šA.n0.f0.ə = *šˈi/unfv = ϣ(ⲉ)ⲛϥⲉ. This, however, has to be considered as a somehow conventional solution, and it has to be kept in mind that the presence of the may be a purely graphic phenomenon, with no phonetic meaning at all. Further studies, on a larger and more specifi c corpus, are needed to better determine the uses and functions of this fi nal in Late Egyptian. §4.5 Special readings for two groups Two groups need special attention, as they appear to have a phonetic value that is in contrast with that which could be inferred from the sheer reading of their consonants. §4.5.1 kꜢ and the development of /a/ in proximity of /k/ It has long been suggested (see e.g. Allen 2013, 25) that in spite of its aliph, the sign = kꜢ was pronounced with a back vowel already in Period 1 and 2, that is before the supposed date of the general /ˈa/ > /ˈo/ vocalic shift, which is usually dated between Period 2 and Period 3 (see above §3.1). The most compelling evidence is provided by the cuneiform transcription kuʔiḫku for Egyptian kꜢ-ḥr-kꜢ = ⲕⲟⲓⲁϩⲕ (S). Allen (2013, 25) has suggested that this pronunciation could be an early attestation of the /a/ > /o/ shift, limited to some specifi c phonetic environments. The present study confi rms these observations, and shed some possible light on the phonetic conditions where such early shift took place.

§4 Group writing -A new model
Various words whose Coptic descendants display a vowel /ⲟ/ are attested in the corpus during Period 1 and Period 2. The group writing form of many of them 40 implies a non-back vowel, thus showing that, as expected, the /ˈa/ > /ˈo/ shift did not take place yet. An example could be ym = "sea", which is spelled as / = *yˈam > ⲉⲓⲟⲙ (I.2, II.2) in Period 1 and 2 and therefore does not display any back vowel, but which does present a spelling with w = U in Period 3, after the shift had taken place, as it appears from the form = *yˈom > ⲉⲓⲟⲙ (Period 3, III.2). The spelling of three other words, however, indicate the presence of a back vowel also in periods when an /ˈa/ or /ˈa:/ would be expected.
To which we can add the abovementioned Cuneiform kuʔiḫku for Egyptian kꜢ-ḥr-kꜢ = ⲕⲟⲓⲁϩⲕ (S).
The use of in the spelling of in Period 1 suggests that such phenomenon was not limited to /a/ > /o/, but aff ected also the shift /a:/ > /o:/, a shift that in normal condition took place only after Period 1.
Even though the evidence is rather scanty, one feature seems to emerge: all the forms characterised by such unexpected back vowel present a velar consonant /k/ just before it. 41 It can therefore be suggested that in general /ˈa:/ > /ˈo:/ after Period 1 and /ˈa/ > /ˈo/ after Period 2 except after /k/, where such shifts may occur earlier, already or even before Period 1, at least in some words. 42 To investigate the reasons for such an early shift in such a phonetic environment is beyond the scope of this study. Some preliminary ob-40 Their full list is given below ( §5. [1][2] and in Appendix A. 41 Which often evolves into ϭ = /kʲ/ in Coptic. 42 The conditional here is due, because three attestations are hardly enough to generalise the phenomenon. servations, however, can be put forward. First, a plain voiceless velar /k/ is not an obvious phonetic trigger for the backing of a following vowel. However, a partial parallel for such a phonetic development can be found in Proto-Khanti, where [ɑ:] > [ɔ:] before velars (Zhivlov 2014, 124, n5). Alternatively, and perhaps more likely, one could assume that such shift was triggered by some additional feature that characterised the consonant k, but which was not recorded in the Egyptian writing. In particular, if the k was realised as a labialised *kʷ or as "emphatic" *kʼ (possibly pharyngealised [kˤ]?), then such a development would look less surprising. 43 Moreover, a secondary articulation of some sort could also help explaining why some of these instances of k evolved into ϭ = /kʲ/ in Coptic, rather than into plain ⲕ = /k/.

Period
Further research, however, is needed to clarify these points.
In any case, it seems that the presence of a consonant k may trigger the backing of a following /a/ or /a:/ already in Period 1, and this has to be considered when analysing the vocalic values of group writing. §4.5.

= dʸ
The group = dʸ is attested in three words in the corpus, and in all of them it appears to transcribe a back vowel. In the case of , such a reading is confi rmed by a later variant from Kawa (Hoch 1994, 252, no.353), where the word ϣⲟⲣⲧ -ḫ₂rd is spelled = ḫ₂0.Ur.d = ḫ₂Urd, which confi rms the presence of a back vowel in this word and suggest a reading dU/Ud for the group = dʸ, and therefore Uʳd for . 43 Backing of vowels after labialised consonants is attested in various languages, while a good examples of that phenomenon in association with "emphatic" consonants is provided by Arabic, where the vowel /a/ is usually realised as /ɑ/ after emphatic consonants. Moreover, emphatic consonants are common in various Afro-Asiatic languages (e.g. Bisang 2006, 80 with refs) and labialised consonants are and probably were present in at least some subfamilies (possibly including Proto-Semitic - Diakonoff 1975, 141) if not even in Proto-Afro-Asiatic itself (Bomhard 1984, 185), and therefore it would not be so surprising if they existed also in Egyptian, at least in some specifi c phonetic contexts.

§4 Group writing -A new model
As said above, the sign is used here to mark a somehow non-standard pronunciation of the sign = d. The use of this alternative group = dʸ to transcribe the syllable d/tU, instead of = tw, which would have been the obvious candidate, could have emerged from the need of preventing ambiguity, as the group = tw fulfi ls already other linguistically distinct functions in the Late Egyptian orthography. 44 The reasons for choosing specifically the group = dʸ, and the origins of its apparent value with back vowel, however, escape me.
For further discussion on this group , see below §11.
All the forms attested in the corpus are discussed in the following paragraphs according to a division by period and by the nature of the stressed vowel in the contemporary pronunciation. Cases involving a /k/ + /a/~/o/ are discussed separately for each period. §5.1 Period 1 In Period 1, 15 words with a stressed non-back vowel are attested. All of them are written without w. In 5 of them, the stressed vowel could be reconstructed both as a non-back /i/ or a back /u/. The spelling of the Egyptian forms, however, clearly points to a nonback vowel. Related Semitic forms are identifi able for one of them, and they support this interpretation (see below §6 for a throughout discussion of Semitic forms related the words of the corpus): 3 of these words are written with a w = U in the actual position of the vowel, in the form c-w + c = cU + c = cU.c. In 4 cases the w = U is written after the consonant following the vowel, that is as c + c-w = c +Uc = c.Uc. In 1 case, w = U follows a cluster r+c, in the form c + ʸr + c-w = Urc = c.Urc. Finally, 1 case is written according to the form c-w + c-w = cU + Uc = cUc.

Egyptian /a/ = A > Coptic ⲟ; Egyptian /a:/ = A > Coptic
The attestations are summarised in the following table:  2   7 of them according to the form c-w + c = cU + c = cU.c. 9 according to the form c + c-w = c +Uc = c.Uc. 4 according to the form c-w + c-w = cU + Uc = cUc. 3 both according to the form c-w + c = cU + c = cU.c and c + c-w = c +Uc = c.Uc or with a cluster r+c, in the form c + ʸr + c-w = Urc = c.Urc. 2 in both the forms c-w + c = cU + c = cU.c and c-w + c-w = cU + Uc = cUc. In 1 case, w = U follows a cluster r+c, in the form c + ʸr + c-w = Urc = c.Urc. The present interpretative model allows to defi ne the Egyptian vocalisations of words for which the Coptic data alone are ambiguous. In particular, Coptic vowels ⲉ = /e/, ⲏ = /e:/, and in some cases ⲓ = /i:/ can derive from both the Egyptian vowels /i/ and /i:/, and /u/ and /u:/, and there is generally no way to determine the right ancestor without external data. 45 Since however /i/ and /i:/ are non-back vowels, while /u/ and /u:/ are back vowels, they are distinguished in group writing transcriptions by the absence or presence of the marker w. Some of these words appear to be Semitic loans, and the comparison with related Semitic forms, when available, confi rms the validity of this approach. As discussed in the introduction, using Semitic data to establish the vocalic value of the groups is problematic for various reasons. Semitic forms, however, can provide precious information to verify the values defi ned through other sources like Coptic, as it is the case here. Naturally, issues concerning dialectal variations in Semitic, as well as the problem that the actual borrowing of a word may predate its fi rst attestation in the Egyptian sources of decades or even centuries, have to be considered. As it appears, however, in almost all of the cases the evidence is very clear, and in agreement with the value suggested by the group writing spelling. Only one case is too ambiguous to be conclusive, and this because the Semitic evidence itself is problematic and contradictory.

Egyptian
Related Semitic forms can be identifi ed for 13 words in the corpus whose stressed vowel can be reconstructed as either /i/ or /i:/, or /u/ or /u:/.
• III.13 "commerce", "associate", "companion" Period 3 ḫAbAr *ḫvbˈi:/u:r *ḫaber Related vocalised Semitic forms: Heb. ḥāber "associate" (Hoch 1994, 241, no. 333). Related non-vocalised Semitic forms: Arb. ḫbr "to negotiate" Notes: it is impossible to say if it was borrowed as /i:/ and then turned into /e:/ within Egyptian, or if it was already borrowed as /e:/. The Egyptian form was clearly borrowed from a dialect where the stress had already moved to the last syllable, like in Hebrew.
• II.45 "oven" Period 2 tArUr *tvrˈi:/u:r(v) *tv(n)nu:r(v) Related vocalised Semitic forms: Akk. tinūru "oven"; Arb. tannūr "oven"; Heb. tannūr "oven"; Syr. tannūrā "oven" (Hoch 1994, 359, no. 351). Related non-vocalised Semitic forms: Ugaritic tnrr "oven" Notes: the fi rst /r/ of the Egyptian form is usually assumed to be due to assimilation to the second /r/ (so Hoch). I wonder however if it could indicate that the word originates from a northern dialect akin to Ugaritic. If we assume a vocalisation *tvnrur for the Ugaritic form, a loan from a similar form with subsequent simplifi cation of the cluster *tvnrur > tv(r)rur could also be a valid explanation.

Coptic ⲁ/0 < A ~/i/ -Semitic non-conclusive
• II.37 "ashes", "cinder", "embers" Related vocalised Semitic forms: Akk. gumāru "burning coal"; Arb. jamra "live coal"; Arm. gūmartā "burning coal"; Syr. gəmurtā "live coal" (Hoch 1994, 301, no. 435). Related non-vocalised Semitic forms: Ugaritic gmr "burning coal" Notes: The Semitic evidence is contradictory and inconclusive, and does not allow to suggest a single common prototype. The Akkadian form is a hapax, and it is likely directly related with the Aramaic one. It is however unclear if it is a loan from Akkadian into Aramaic, or from Aramaic into Akkadian (Abraham and Sokoloff 2012, 32). The Arabic form is related, but displays a clearly diff erent vocalic pattern. The Egyptian form could derive from yet another unattested Semitic form characterised by a vowel /i/. §7 Synchronic analysis The distribution of groups with Ꜣ = A, with Ꜣ = 0 and of isolated consonantal signs for marking stressed non-back vowels does not seem to reveal any clear pattern.
One cannot exclude that further studies may reverse this observation, but on the basis of the current evidence it looks like groups with Ꜣ = A/0 and single consonantal signs without it are functionally equivalent, and both can transcribe consonants followed by stressed non-back vowels (i.e. -cˈA-), consonants followed by unstressed vowels (i.e.cv-), which may have already been realised as /ə/, as well as consonants in fi nal position or consonants followed by other consonants (i.e. -c# and -c.c-).
By contrast, some tendencies and patterns seem to emerge from a careful analysis of the forms displaying a stressed back vowel.
In particular, the evidence suggests that the use of c-w = cU or c-w = Uc may correlate in some way with the position of the stressed vowel within the word and in relation with the surrounding consonants.
Four specifi c environments, associated with specifi c spelling sequences, are identifi able: The meaning of these patterns is diffi cult to assess, although some observations are possible. First, it is clear that these four environments may overlap, and therefore multiple valid spelling may exist for structurally similar words, or even for the same word. For instance, monosyllabic words of type #cUc# may appear either written with w = U after the fi rst consonant, thus as #cU + c(A) = #cU.c-, such as = bU.n₁r0 = bUⁿr = *bˈi/ uⁿr (II.8) and = yU.m0 = yUm = *yˈom (III.2), or may also appear with w = U written after the second consonant, thus as c(A) + Uc# = -c.Uc#, such as = ḫ0.Ur = ḫUr = *ḫˈur(rv) (I.15; II.30) or = k0.Up = kUp = *kˈop (III.20). In the fi rst case, the spelling may be justifi ed as a manifestation of environment 1), while in the second it looks like a manifestation of environment 2). Other words can also be spelled in multiple, equivalent ways, that can be explain according to either one or another of the environments 48 §7 Synchronic analysis above. For instance, the spelling = ʕA.n₁r0.Ur = ʕAⁿrUr = *ʕvⁿrˈo:rv can be analysed as an instance of environment 2, while the alternative spelling / = ʕA.n₁rU.ʸr0 = ʕAⁿrUr = *ʕvⁿrˈo:rv can be associated with environment 4a).
The reasons for favouring one spelling over the other are not clear, although the fact that most words seem to be written rather consistently in only one specifi c way suggests that the phenomenon was not totally random. Further studies, however, would be needed to identify the underlying rules, if they existed.
Right now, we can only observe that in the case of the environment 1), it is likely that a w = U after the fi rst consonant can only be pronounced there where it is written, because as far as we know, pre-Coptic Egyptian words could not begin with a stressed vowel, but only with a consonant (Loprieno 1995, 37, 40).
Moreover, the third environment is characterised by the presence of a w = U in two consecutive groups, such as in the case of = tA.rU.Ur = tArUr = *tvrˈu:r(v) (II.45). It appears that in most cases, such groups share the same vowel, and at the same time all back vowels between identical consonants attested in the corpus are spelled in this way. 47 This consistency in spelling can hardly be accidental. One can thus suggests the presence of an orthographic rule according to which a back vowel between two identical consonants should be spelled by reduplicating the corresponding c-w group. However, sequences with back vowel between diff erent vowels, namely -c₁-w + c₂-w-, are also attested. In these cases, the reasons for such spellings are unclear, but it is reasonable to assumed that, like the previous ones, these sequences should also be interpreted as -c₁-w + c₂-w-= -c₁U + Uc₂-= -c₁Uc₂-. Some caution, however, is due: since the nature of the unstressed vowels remain out of our reach, one cannot exclude a priori that, at least in some cases, such spellings were indeed meant to transcribe a -c₁U + c₂U-= -c₁Uc₂Usequence.
Finally, the data for Environment 4 strongly suggest that within a word, a CU group is by far more likely to transcribe a CU sequence rather than a sequence UC. This seems to be especially true when such group corresponds to the second consonant of the word (i.e. a #cv.cU-context): in 7 attestations over 9, such groups transcribe CU sequences. It would be tempting to correlate this observation with Environment 1, and to suggest that a CU group in second position, i.e. in a #cv.cU-syllabic sequence, can be assumed to represent CU rather than UC, because a syllabic sequence #cU.c-would already be covered by Environment 1, making a spelling #c0.UC-= #c.UC-for such syllabic sequence rather superfl uous.
However, the presence in the corpus of two words, namely = ḫ0.Ub.sA = ḫUb(A)sA = *ḫˈi:/u:bvsv (II.27) and = b0.Ur.y(U)/(U)y = bUry(U)/(U)y = *bˈoryv (III.5), which do use the spelling #c0.UC-to transcribe a sequence #cUc-, calls for attention. The name of the city of Ugarit, usually written in the Egyptian texts as (see Gauthier 1925Gauthier -1931 for examples) and to be read as ʔ0.Uk.rʸA.t(A) = 47 The only exception is ṯṯ = "sparrow" (II.46), which is spelled in two ways, namely , thus according to the rule, but also , thus in a irregular way. In light of what just discussed, the second spelling may be a scribal mistake.

ʔUk(A)rAt(A)
, can also be cited here. The reading of the name of Ugarit is confi rmed by the cuneiform sources, where it is spelled u₂-ga-ri-it/tv = ʔugarītv, and there is no possibly ambiguity: the back vowel /u/ was located after the fi rst consonant, not after the second one.
The reasons for such spellings, if any, remain unclear. It is possible that they are the direct or indirect consequence of specifi c and consciously applied orthographic rules. If so, however, the rationale behind such rules remains to be discovered, although it would probably have to be searched in the word-based rebus nature of the group writing and in the w-extended orthography discussed above. It may also be that, perhaps, some consonants were somehow inherently incompatible with the marker w = U, and therefore a sequence involving such consonants followed by a back vowel could be transcribed only through a c(Ꜣ) + cw = -c0.UC-= -c.UC-spelling.
For instance, neither in the corpus nor in Hoch 1994, 510 there is any reliable attestation of the use of groups *ḫw to transcribe the sequence ḫU. 48 Rather, when such sequence does occur, it is transcribed through a ḫ(Ꜣ) + cw = -c0.UC-= -ḫ.UC-spelling. The words ḫ .Ur = "road", "street", "quarter" (I.14;II.29; III.14) ḫ .Ur = "Syrian" (I.15; II.30) and ḫ .Ub.sA/0 = "lamp" (II.27) in the corpus are good examples of this. Alternatively, perhaps some sequences of consonants were strictly identifi ed with specifi c words, within the rebus-nature of the group writing, and therefore were just spelled in such way because they were learnt as such.
The evidence provided by the corpus studied here is not enough to clarify these aspects, and further research is needed.
The potential theoretical ambiguity rising from the coexistence of these possible double reading of the same groups is evident. However, if one looks at the specifi c transcriptions of these words, one realises that in fact actual ambiguity is relatively rare.
Just to mention a few examples, the word = *svrpˈot(v) (III.16) could in theory be read both sA.ʸr0.pU.t(A) = sArpUt(A) or s0.Uʸrp.t(A) = sUrpt(A), but the latter option can be safely excluded because the cluster of three consonants rpt, implied by the sign , would be incompatible with the rules of Egyptian phonology as we know them, and would have probably been simplifi ed in some way if it came from a foreign word. The same stands true for a word like = *ḏvḥˈi/urtv (II.49), where a reading ḏA.ḥU.ʸr0.tA = ḏAḥUrtA is the only possible one, because a reading ḏ0.Uḥ.ʸr0.tA = ḏUḥrtA would generate an unlikely cluster ḥrt. In the case of = *ḥvlˈi:/u:lwv (I.13), instead, the only possible reading is ḥA.rU.r0.w₁A = ḥArUrw₁A, because a reading ḥ0.Ur. rA.w₁A = ḥUrrAw₁A would imply a geminated /r/, As far as we know, however, Egyptian orthography did not spelled out geminated consonants and wrote them like normal, simple ones. Therefore, a gemnated /r/ would have likely not been spelled out with a double r, as this form would require (the r of and that of ).
48 The only attestation recorded by Hoch, , is not a group writing spelling but rather a case of w-extended orthography, and should be interpreted as prḫ+w = prḫ+U = pUrḫ = *po:rḫ, from which Coptic ⲡⲱⲣϣ (S) = po:rš.

§7 Synchronic analysis
As it appears, the spelling of other syllables around the c-w group is often enough to solve the ambiguity, often thanks to the presence of a . Moreover, the presence of a may trigger the application of orthographic conventions also in other contexts. For instance, cases like = *mvḫ₂ˈi:/u:rv (II.16) may suggest that the presence of a meant to modify the value of a consonant -like in the case of = ḫ₂-was incompatible with the presence of a w = U in the same position. This idea seems to be supported by the absence of any group among the hundreds of words studied by Hoch (1994, 510). This could hint to the existence of an orthographic rule according to which a string C₁UC₂ where C₁ had to be spelled with a sign, had to be transcribed with a sequence cʸ(Ꜣ) + cw = Cʸ(0). UC, such as = ḫʸ (0).Ur = ḫ₂ˈUr. These considerations may suggest that, in fact, resolving ambiguous spellings may have been a major function of the sign , and this especially in those cases that cannot be explained as in §4.2 above. The exact rules regulating such possible use of , however, are diffi cult to defi ne on the basis of the corpus used in this study. The examples are just too few to try to extrapolate any general pattern from them. A more specifi c study, looking more systematically at the occurrences of and based on a wider corpus, including also forms unattested in Coptic, may help in clarifying this issue.
Since some of the words of the corpus are attested in diff erent periods, the model presented in this book can be verifi ed also from a diachronic perspective. In particular, there are 21 words attested in more than one period, and the main phonological changes postulated for the Egyptian vowels can be observed taking place through them. In particular: In the Coptic dialects in which it is attested, such as Sahidic and Bohairic, the shift /a/ > /o/ is usually assumed to have taken place at some point around or after 1000 BCE (see above §3.1). This development can be seen in various words of the corpus. Two of them are attested in all the periods, while one is attested only in the fi rst two periods, and one only in the fi rst and last. All these words are spelled without any w in Period 1 and 2. This is what one would expect, as at that time their stressed vowel was still the non-back vowel /a/. In the third period, instead, all these words are written with w, which implies that the presence of a back vowel. This, again, is exactly what one would expect, as by that time the stressed vowel /a/ is assumed to have turned into /o/.  It is usually assumed that the Coptic vowel ⲓ = /i:/ derives either from an earlier vowel /i:/, or from a vowel /u:/ when the stressed vowel is in proximity of /r/ or after a pharyngeal (Loprieno 1995, 48;Peust 1999, 231). Since neither of these conditions apply to the case of , it can be assumed that Coptic ⲓ = /i:/ < /i:/. The word is written without -w in both Period 2 and 3, while it is not attested in Period 1. This spelling implies a stressed non-back vowel, as expected in the case of Coptic ⲓ = /i:/ < /i:/.
The same stands true also for the word qrr: although the Egyptian spelling display two = r, the Coptic form suggest that this consonant was actually pronounced /l/. The vowel can thus be reasonably reconstructed as /i:/. The Coptic vowel ⲉ = /e/ is assumed to derive either from /i/ or from /u/. The corpus contains one word which displays a vowel /e/ in Coptic and a transcription implying a non-back vowel in Egyptian. The word is attested in Period 2 and 3, and both transcriptions imply non-back stressed vowels, as expected in the case of /i/ > /e/. Since the group writing distinguishes only between non-back and back vowel, but not between /i/ and /e/, it is not possible to use these forms to confi rm that the change /i/ > /e/ took place between Period 2 and Period 3. As said above, it is usually assumed that the Coptic vowel ⲓ = /i:/ may come from an earlier vowel /u:/, if next to /r/ or after pharyngeal. One word of the corpus, = ḫr = ϩⲓⲣ, attests this development. It is likely that = ḫr comes from Semitic ḫur(ru) (Hoch 1994, 247-8, no. 343;Loprieno 1995, 48), and thus displayed an original /u(:)/ vowel. The word = ḫr is written with w = U in all the three periods, which on the one hand confi rms the original back nature of the stressed vowel and on the other suggests that the shift /u:/ > /i:/ took place at some point after Period 3. This is a crucial observation for the understanding of such phonological development, as discussed more specifi cally below ( §10). As for , no external evidence can be used to infer the nature of the original vowel. The Egyptian transcriptions however consistently point to a back vowel, and the fact that no change is observed between the spelling of period I and period II confi rms that the shift /u:/ > /i:/ took place at a later date. Three words in the corpus attest the second development. In the case of , the /u:/ nature of the stressed vowel is supported by the Semitic parallels ( §6). In the case of and / , instead, it can be inferred by the spelling with w, suggesting the presence of a back vowel.
As said above ( §3.1), the exact date of the development /u:/ > ⲏ is not clear. Moreover, as mentioned, it seems that Coptic ⲏ corresponded to two diff erent non-back phonemes, namely an unrounded one -usually described as /e:/-and a rounded one -usually described as /ø:/-, although it is not clear how such Coptic phonemes would related with the earlier /i:/ and /u:/. In any case, what emerges from the word discussed here is that one of the ancestors of Coptic ⲏ remained a back vowel at least until Period 3.

§8 Diachronic analysis
Coptic ⲁ = /a/ < /e/ < Egyptian /u/ The development Egyptian /i/ > /e/ > Coptic /a/ = ⲁ is attested by one word, whose spelling suggests the presence of a non-back vowel during all the three Periods. Since the group writing orthography does not allow to distinguish between /i/, /e/ and /a/, being all non-back vowels, it is not possible to determine when these shifts took place.

§8 Diachronic analysis
The correspondence between the readings of these words in various periods and the expected evolution of the vocalisation validates the interpretation of the group writing presented in this study. At the same time, it also confi rms and in some cases allows refi ning our understanding of the evolution of the Egyptian vocalic system. §9 Statistical analysis §9.1 Theoretical background Given the intrinsic leeway of a system that marks only two vocalic classes, and in which groups encoding back vowels can be read both as cU and as Uc, it is reasonable to wonder if the results presented above may actually be due to mere chance. This issue can be rephrased as a very specifi c probability question: with the present rules, what is the probability that a group with -w will always correspond to a stressed back vowels (either through a reading cU or a reading Uc), while a group without -w will always correspond to a stressed non-back vowel? The statistical problem underling this question is diffi cult to solve in a mathematical way, because multiple variables 49 play a role in the outcome. This problem, however, can be eff ectively addressed with an empirical statistical approach based on the so-called Monte Carlo methods. 50 Basically, Monte Carlo methods rely on repeated random sampling to obtain empirical estimations of the probability of a given event. For instance, let us assume that we have a deck of cards. We draw 10 cards from it, and we fi nd out that all of them are red. What is the probability of such an outcome occurred by mere chance, rather than because, for instance, someone manipulate the deck and put only red cards at the top? A possible Monte Carlo approach to solve this problem would be to perform multiple trials, each consisting in re-shuffl ing the deck and re-drawing 10 cards, recording each time how many black cards and how many red cards have appeared.
If enough trials are performed, the observed frequencies of the various outcomes will represent a good empirical approximation of their actual probability. If we organise the outcomes in a chart, starting from a "10 black cards" outcome and progressively moving toward a "10 red cards" one, the frequencies and therefore the probabilities of such outcomes will be distributed according to a so-called Gaussian or bell curve. The outcome "5 red, 5 black" will have the highest probability, 51 and the other outcomes will be distributed around it in a decreasing way, tending to zero toward the two edges of the chart.
A similar Monte Carlo approach can be used to assess the probability, and therefore the likelihood, that the matches between the reconstructed vocalisations and the readings of the group writing spelling suggested in this study may be due to mere chance.
The problem can be framed as follow: the number of matches between reconstructed vocalisations and suggested readings (which equals the very number of words in the corpus, as all of them can be regularly explained according to the current model) has to be compared with the number of matches that can be obtained between the same vocalisa-49 For instance, one has to consider that in Egyptian the frequencies of stressed back and non-back vowels are diff erent, and that such frequencies vary depending both on the syllable in which they appear and, more in general, on the specifi c phonemic inventory of each period. 50 For a good, general introduction to Monte Carlo methods, see Fishman 1995. 51 Since there is an equal number of red and black cards in a deck, a 50%-50% outcome is the most probable one. tions and the same readings when they are paired at random, over multiple trials. A match is obtained every time that, in such randomly generated pairs, the group writing spelling can be explained as a transcription of the paired vocalic pattern, according to the rules suggested for the model presented here. If the results presented in this book were truly accidental, one would expect cases in which all group writing spellings match the corresponding randomly paired vocalic patterns to be relatively common, to be a recurrent outcome over multiple trials, and therefore to have a rather high probability to occur by chance. By contrast, if such cases happen to be overall rare, then this would strongly suggest that the results presented in this book are statistically signifi cant and are likely not due to chance. This, in turn, would strongly suggest that the interpretative model as a whole is valid.
Such a Monte Carlo test can be easily implemented into a simple computer program that automatically performs n trials and counts the outcomes. The test itself can be built and encoded in various ways. A particularly easy one consists in creating an array_1 with all the vocalic patterns attested in the corpus, an array_2 with all the attested group writing spellings, and a match_table listing the various pairs of vocalic pattern -group writing spelling that should be considered as valid matches. The program will then perform multiple trials in which the items in array_1 and array_2 are fi rst shuffl ed and then randomly paired. The resulting pairs will then be assessed against those recorded in the match_table, and the valid matches will be counted for each trial.
It is important to note that the corpus is composed of words displaying diff erent numbers of syllables, and which therefore have vocalic patterns and group writing spellings of diff erent lengths. In particular, most of the words of the corpus have either 2 or 3 syllables, while only a few are longer. This aspect has to be considered while performing the Monte Carlo test, because it is clear that each vocalic pattern should be paired only with spellings of the same length. It would not make much sense to compare, for instance, a vocalic pattern with two syllables, with a spelling composed of three groups. In order to deal with this issue, the corpus has been divided into batches of words of the same length. The Monte Carlo test has then been performed independently on the batches of disyllabic and trisyllabic words, which constitute the majority of the corpus. Longer words have been ignored, because they are too few to be meaningfully tested. I have implemented the procedure just described as a Python script, which is freely available in my github repository ( https://github.com/MKilani/LingAeg_group_writing_Monte_Carlo_test ).
In order to compare them, vocalic and spelling patterns need to be transcribed and encoded in a coherent way. To do so, a few factors need consideration. First, obviously, only a distinction between non-back and back vowels is needed in the transcriptions of the vocalic patterns, because only such distinction is refl ected in the group writing spellings. Since only stressed vowels have been studied in the this book, unstressed vowels can be ignored in the encoding of the words and in the subsequent assessment. It has to be noted that in some cases, it is impossible to reconstruct with any certitude if the stressed vowel was a non-back or a back vowel. This is the case, for instance, of the ancestors of Coptic ⲉ, which could be either /i/ or /u/ in Period 1 and 2. In these cases, the vocalic patterns must be transcribed in a way that can be matched both with the spellings implying a stressed 61 §9.1 Theoretical background non-back vowel and with the spellings impying a stressed back one. Finally, as discussed above, preconsonantal r-and n-clustered with the following consonants. This means that from the point of view of group writing spellings, the sequences r-and n-+ consonant behaved as single consonantal segments. Therefore, such clusters should be treated as single consonants in the transcription of the spelling patterns displaying them.     The graphic representations of the probabilities of random matches are provided in Appendix B.
The results are clear: in all the cases, the probability of obtaining by chance a valid match for each of the pairs is extremely low. In many cases, it is so low that it cannot even be precisely estimated. These sets of data unequivocally suggest that it is statistically very unlikely that the results presented in this book are due to mere chance, and therefore they provide strong additional evidence of the validity of the system. §10 The development of Coptic ⲏ As mentioned above ( §3.1), the development /u:/ ~ /i:/ > ⲏ is a thorny issue in Egyptian historical linguistics. The evidence presented in this book provides some new relevant data. What emerges is that during all the three periods, Coptic ⲏ could correspond to both a non-back and a back vowel. This suggests that either these two vowels never fully merged, as suggested by some scholars (see Peust 1999, 228-30), or such merging took place after Period 3. This data are in agreement with what emerged from the preliminary study of the w-extended orthography, which also suggested the presence of two distinct vowels, a back and a non-back one, in correspondence of Coptic ⲏ in texts dating to Period 3 (Kilani 2017a). It is also worth to observe that the word = hm (II.24), attested in the corpus only in Period 2 and spelled with , which implies a pronunciation *hˈu:mv with /u:/, is instead spelled without in Pap. BM EA 10474 (Teachings of Amenemope) 27.3,4, which dates to the 26th Dynasty (Laisney 2007, 6), thus suggesting a non-back stressed vowel, which may imply a pronunciation *hˈe/ø:mv. This could therefore indicate that the /u:/ > ⲏ (= /e:/ or /ø:/?) shift took place after the end of the 22nd Dynasty (i.e. after Period 3) but before the 26th Dynasty.
The specifi c nature of such back vowels cannot be specifi ed on the basis of the evidence emerged from this book. It can only be said that a non-back and a back vowel were involved, but it is not possible to say if such vowels were maintained as /u:/ and /i:/ during the three periods, or if they underwent some minor shift that did not change their back and non-back nature, such as /i:/ > /e:/. §11 The group -additional observations As explained above, the few examples in the corpus show that the group encodes a back vowel and can be read as dU or Ud. This assumption can be confi rmed on the basis of other attestations in toponyms, personal names, and words that have no direct descendant in Coptic and which, therefore, are not part of the corpus. I discuss them in detail here below.

Attestations
Group Writing Transliteration Reconstruction Prototype Period 3 ( ) Eg.: Gloss.  The spelling for the name of the city of Armant is attested once, in the Onomasticon of Amenope (no. 332). The name of the city is attested in Coptic as ⲣⲙⲟⲛⲧ = (ə)rmˈont (S) and ⲉⲣⲙⲟⲛⲧ = ərmˈont (B), in Greek as Ἑρμωνθις = ermōntʰis and Ἑρμονθις = ermontʰis and in Latin as Hermunthus, Hermonthes and Hermonthis. 52 The name can be analysed as = "the city Ôn" and = "the God Montu". The Coptic, Greek and Latin forms of this toponym suggest the presence of a stressed back vowel. The pronunciation *mˈont(ə) can thus be reconstructed for Period 3. The group writing spelling = m0-Und-t0 = mUnd/t perfectly refl ects this pronunciation. This form provides a good example of = preconsonatal n being clustered with the following consonant, in this case d, thus resulting in = n + d-w = nd-U = Und (see above §4.3).
All these forms present a sequence dōd or dud, which the Egyptians transcribed as = dU-Ud = dUd.
• "skilled", "knowing", "knowledgeable" The word appears in Papyrus Anastasi I in the expression = "skilled/learned scribe " Hoch 1994, 58, no. 64. The meaning and Semitic prototypes of the two words are clear, but their vocalisation has been a subject of debate. Hoch compares = ṯUpAr 54 <?> with Biblical Hebrew sōfer = "scribe" (from a participle *sōpir), and he correctly observes that the vocalisation of the Egyptian form, with a back vowel in the fi rst syllable, points to a coastal North-West-Semitic dialect (Hoch 1994, 58). However, he is then puzzled by the vocalisation of , 55 because according to his system it does not show any trace of the expected /a:/ > /o:/ shift. He thus concludes that either the spelling of was inaccurate regarding the vocalisation, or the scribe mixed forms from two diff erent dialects, one that underwent the Canaanite shift and one that did not. This problem disappears if we consider = dU/Ud and therefore we interpret as y0-Ud-ʕA/0 = yUdʕA/0. This spelling implies the presence of a back vowel, and therefore perfectly refl ects a North-West-Semitic prototype *yōdiʕ = "knowing", "who knows", hence "skilled" (active participle -cf. Biblical Hebrew yōdeaʕ), which is what we would expect in a dialect that underwent the shift /a:/ > /o:/. It thus appears that the new reading suggested here not only provides a likely interpretation for this spelling, but it also elegantly solves the issue noticed by Hoch by showing that both = ṯUpAr from *sōpir or the like, and = yUdʕA/0 from *yōdiʕ or the like come from a same single dialect that underwent the shift /a:/ > /o:/.
• "stand fi rm", "hold one's ground" The word is attested only once in the Egyptian sources, in Period 2, in the expression jry=k ʕAmUd m pꜢ nhrn = "you stand fi rm in Naharin".
is certainly related with Semitic √ʕ-m-d = "to stand fi rm" -see Akkadian emēdu = "to stand (near) by", "to lean on"; Arabic ʕ-m-d = "to approach", "to support"; Aramaic ʕă mad = "to place (G-stem)", "to place (D-stem)"; Biblical Hebrew ʕāmad = "to stand (up against)". The 54 He reads it, according to his system, as ṯu=pi ₃ =ra Hoch 1994, 364. 55 He reads it as ya=di=ʕa, according to his system. root is well attested in various North-West-Semitic languages, but it is common as a verb only in Hebrew. The Egyptian construction in which this word appears would require an infi nitive, and in fact the vocalisation of , with a back vowel in the second syllable, could indeed correspond to a North-West-Semitic infi nitive in a language that underwent the Canaanite /a:/ > /o:/ shift, as attested by the Biblical Hebrew infi nitive ʕāmōd.
• "Dor" (toponym) Period 3 dUr *dˈUr *do:r/*doʔr Eg.: LES 5 1.8 The city of is attested only once in the Egyptian sources, in the tale of Wenamun. It is usually identifi ed with the city of Dor, about 30 km South of Haifa. Its name is attested in Hebrew as doʔr (Joshua 17:11) and dōr (1 Chronicles 7:29), and in Assyrian Akkadian as ᵁᴿᵁdu-uˈ-ru (Gilboa and Sharon 2016, 241). The back vowel present in all these forms is refl ected in the use of the group = dU, which in Period 3 stands for /do/, /do:/ or /du:/.
• "amorous", "lustful", " lascivious" The word and the name have long been recognised as being related with the Semitic root √d(-w)-d = "to love (et cetera)" -see Biblical Hebrew dōd = "lover"; dōdīm = "love-(making)"; Ugaritic ddm = "love"; Aramaic dōda = "lover"; Akkadian dādū "love-making". Since the group has usually been read as dy, until now these words have been interpreted as dydy, didi and the like. However, as Hoch (1994, 379) points out, such a vocalisation is grammatically problematic and has no good parallel in any attested Semitic form. By contrast, reading the group as dU/Ud, and therefore the sequence as dU + Ud = dUd solves the issue. The word = dUd, can then be compared with Hebrew and Aramaic dōd(a) = "lover" (Hoch 1994, 379). As for the name = dUd, it can be compared with ᵐDu-u(₂)-du, attested in the Amarna letters as the name of an Egyptian offi cial (EA 158:1,5,12,34;EA 164:1,10,16;167:28;EA 169:16), This name is in turn clearly related with Biblical Hebrew dāwid. In fact, both the name attested in the Amarna letters and the Egyptian = dUd could be interpreted as The group presents some peculiar characteristics and therefore deserves a specifi c discussion. This group is not attested in the corpus, as none of the nouns presenting it has a sure direct descendant in Coptic. , however, does appear in the Period 3 spelling of a toponym, = "Sile", which is later attested in Latin, Greek and Coptic. This spelling is recorded in the version of the Onomasticon of Amenope preserved in the papyrus known as the Golenischeff Onomasticon (Gardiner 1947, Vol. I, 27-9). Eight additional words characterised by a fi nal are present in this papyrus. Three of them end with a sequence , which suggests they may be characterised by vocalic patterns comparable to that of . None of these words, however, survives in Coptic, and therefore their vocalisation cannot be externally verifi ed.
The group appears in the Onomasticon also at the end of a few more words spelled in traditional orthography. Some of these words do survive in Coptic, and therefore they provide crucial information about the value and functioning of this group. Since the Onomasticon of Amenope provides a small but rich and coherent corpus, all words displaying a fi nal attested in it are listed and singularly discussed in detail here below, at the end of this section. §12.  56 In their case, therefore, it is not possible to exclude a reciprocal, Greek-Coptic infl uence, and if so it is not possible to defi ne the direction of such infl uence: the city is obviously in Egypt, and therefore it is likely that the Greek spelling refl ects an Egyptian pronunciation. The Coptic spelling may also refl ect the same pronunciation, and may thus truly be an indigenous rendition of the same name, but it could also be a back formation shaped on the Greek form. 57 There is hardly any way to solve this issue. However, even leaving aside the Coptic form, a few observations can be made: both the Greek form Σελη and the Latin Sile suggest a diff erence between the fi rst and the 72 §12 The group second vowel of the name. The Greek form suggests that such diff erence resides in its prosody, while the variation between Latin Selle and Sile may hint at the fact that the last vowel was perceived unambiguously as e, while the fi rst one may have been perceived as less defi ned and somehow intermediate. By combining these considerations, one can reconstruct the pronunciation underling these forms as *səlˈe:, which in fact is also the pronunciation that underlies the Coptic spelling. This reconstruction is confi rmed also by the Akkadian spelling of the name of this city, which is attested in an Amarna letter (EA 288.46) as URU Sí-lu-ú = Silû. This latter spelling can be assumed to represent a Period 1 pronunciation *si/əlˈu:, which can be expected to have regularly evolved into a Coptic and later pronunciation *səlˈe: , with stressed /u:/ > /e:/ = ⲏ after Period 3. Since the spelling dates to Period 3, it refl ects the pre-shift pronunciation. The fi rst part is a rather straightforward transcription 58 for a pronunciation *si/əlˈu:. The fi nal , instead, deserves more attention. Let start by leaving aside for a moment the value of the sign , and by transcribing it just as ʾ. According to the model presented above, the group can be read as ʾ + U = ʾU/Uʾ. Overall, therefore, the spelling can be understood as ṯA + rU + Uʾ = ṯArUʾ = *ṯəru:(ʾ).
This considered, I think that the group as a whole can be interpreted as indicating a fi nal stressed back vowel. Conceptually, this group may be a fi nal equivalent of the initial group , which indicates the presence of an initial non-back or unstressed vowel possibly preceded by a glide or glottal stop (Loprieno 1995, 38 n38, 247;Allen 2013, 32). Similarly, the fi nal stressed back vowel indicated by the group may or may not have been followed by a fi nal glide or glottal stop -hence the transcription with ʾ here above. This last point deserves some discussion. According to the currently accepted models, Late Egyptian should not have words ending in open stressed syllables, that is there should be no word ending with a stressed vowel not followed by any consonant (Loprieno 1995, 39-40). Words ending in stressed vowels, however, do exist in other languages, including North-West-Semitic ones, and they may have thus entered Egyptian as loanwords. At that point, two scenarios were possible: either Egyptians adapted such loanwords to their native Egyptian phonotactics, and some additional element -possibly a glide or a glottal stop-were introduced after the last vowel in order to close the respective syllable, or Egyptians adopted these words as they were, and through them they introduced new, irregular prosodic patterns into the Egyptian language. In the case of , the evidence points to the latter scenario. In particular, we know that this toponym is attested in Egyptian at least since the reign of Thutmose III in Period 1 (Urk. IV 647.11). We also know that the presence of a η/ⲏ in the Greek and Coptic forms suggests that the last vowel was a long vowel /u:/. According to the current models, in Period 1 such vowel could occur only in an open syllable. 59 The evidence from before Period 3 confi rms this interpretation: this toponym is consistently spelled as in both Period 1 and Period 73 §12.1 The group -interpretation 2. 60 These spellings can be interpreted as ṯA + rU = ṯArU = *ṯəru: and as expected they show no trace of any glide or glottal stop after the fi nal U = u:. In the case of , therefore, the of the group seems to have no etymological phonetic value.
Moreover, the group can be used in combination with the group in what looks like a set sequence . The most interesting example is the word (LEM 5 12.2 -Hoch no. 467), which does not have a Coptic descendant but which is clearly a Semitic loanword related with Biblical Hebrew kinnōr. This word shows that the sequence has to be read as nU, 61 because the resulting reading = kAn-nU-Ur = kAn(n)Ur is the only possible one that is both in agreement with the Semitic forms 62 and internally coherent. 63 Moreover, this word suggests that the sequence , and therefore the group , encode for a plain back vowel, because in this word there is no place where a glide or glottal stop could be expected.
The sequence is attested in various words, usually at their end. For instance, the words , and appear in the Onomasticon of Amenope. Neither of them has a Coptic refl ex, therefore their vocalisation cannot be directly reconstructed. On the basis of what has just been discussed, however, it can be inferred that they ended with a stressed back vowel in an open syllable, as further discussed in the list below ( §12.2).
These observations seem to suggest that the group was indeed transcribing just a plain back vowel.
At the same time, however, there is evidence showing that this same group was also used there where a fi nal glide could be expected because of etymological reasons. In particular, the Period 3 vocalisation of the word = "mountain" can be reconstructed as *ḏˈow, on the basis of Coptic ⲧⲟⲟⲩ (S) = tˈow. In this case, therefore, the group seems to transcribe the sequence -ow, which was perhaps perceived as -oʷ, that is as a back vowel -o followed by a glide -ʷ.
The fact that in some cases the fi nal stressed syllable indicated by the group may have been closed, and therefore a glide or a glottal stop may have been present, at least historically, is suggested also by another observation.
In the same way as the group could be added to word written in standard orthography to indicate a vocalic prefi x (see e.g. Junge 2005, 97-8), it seems that the group could be added to words written in standard orthography to indicate, I think, a fi nal stressed back vowel. The Onomasticon of Amenope presents multiple such examples. As it appears from the list below, most of these words have two characteristics in common: 60 With some variation in the classifi ers, see Gauthier 1925Gauthier -1931, VI 67 for attestations. 61 As already suggested by others scholars, see e.g. Edel 1966;Hoch 1994, 508. 62 Either by representing a pronunciation *kin(n)ˈo/ur directly corresponding for instance to Hebrew kinnōr, or by refl ecting a pronunciation *kin(n)ˈor deriving from an earlier **kin(n)ˈar which in turn would directly correspond to an earlier Semitic form kinnār. 63 As the fi nal -n of matches the initial n-of = nU, and the U of = nU matches the U of . 74 §12 The group 1) most of them derive from Middle Egyptian roots which are usually assumed to have ended with a weak consonant -j / -w or with -Ꜣ. 2) for those words that have a Coptic refl ex, the corresponding Period 3 forms can be reconstructed as ending with a stressed back vowel, often followed by a glide.
The words ṯsʾ = "chief" and jqdʾ = "builder" are attested in Coptic only in their absolute state (ⲉⲕⲱⲧ and ϫⲟ(ⲓ)ⲥ respectively), while the Egyptian forms discussed here are likely in the construct state, as they are clearly the fi rst elements of genitival constructions. As discussed below, a direct comparison is therefore not possible, as the forms may be diff erent.
It thus appears that the group can be used to represent both plain fi nal stressed back vowels, and fi nal back vowels followed by some form of glide or etymological glottal stop. Two parallel scenarios can be put forward to explain this situation. A fi rst possibility is that the Egyptians did not conceptualise such fi nal glides and glottal stops as full consonants, and therefore they perceived fi nal stressed back vowel with and without them as equivalent, at least from a writing perspective. The second possibility, instead, is that the phonotactics of Egyptian strictly required fi nal stressed syllables to be closed, and words that did not comply to this rule were automatically extended with such a coda. Such codas were then indicated with the group , at least in Period 3. It seems to me that the fi rst scenario is more likely, but I do not think that, for now, the evidence available is conclusive.
Another word which is relevant for the current discussion is = "stormcloud", "storm", no. 10 in the Onomasticon of Amenope. A semi-reduplicated form also exists (see attestations in WB and TLA), and survives in Coptic as ⲕⲗⲟⲟⲗⲉ, klˈoʔlə = "cloud". The fi rst, short form is well attested in Egyptian in various periods, and from the comparison of such variants its vocalisation can be safely reconstructed as *qvrˈa: > *qvrˈo:. The stressed fi nal vowel is clearly long and therefore the fi nal syllable probably open, because the shift from non-back to back vowel seems to have taken place between Period 1 and 2 (see below for the relevant evidence). By contrast, on the basis of Coptic, the vocalisation of the long form can be safely reconstructed as *qvrˈaʔrv(ʔ) > *qvrˈoʔrv(ʔ) -with r = Coptic ⲗ. A few intriguing observations stem from these two forms and from their comparison. First, it appears that the basic root of the word was √q-r-j, with a fi nal weak consonant, as demonstrated by the early attestations of the short form spelled with a fi nal -(e.g. Middle Kingdom -see WB and TLA), and by the Coptic refl ex of the longer form, where such weak consonant both survives in the middle of the word as a glottal stop and is implied at the end by the fi nal -ⲉ. It is thus clear that the longer form originated from a partial reduplication of the root according to a pattern √c₁-c₂-c₃ > c₁vc₂vc₃c₂vc₃, that is √q-r-j > qvrvjrvj. The stress falls in both forms on the second vowel, which can be reconstructed either as a short or as a long /a/. This identity in the quality of the vowel strongly supports the validity of these reconstructions.
Moreover, as said, the vowel of the short form is long, which suggests that the fi nal syllable was open. This, in turn, indicates that at some point the fi nal weak consonant 75 §12.2 The group -Attestations in the Onomasticon of Amenope must have completely disappeared, 64 thus triggering the lengthening of the fi nal vowel. This process must have taken place both at a time when the phonotactic rules stressed close syllable => short vowel and stressed open syllable => open vowel were still actively operating in the language and before the New Kingdom, as the New Kingdom spellings record the expected development of a long vowel. This is a crucial observation, for multiple reasons. First of all, these attestations show once again that the group can be used to mark a fi nal stressed long back vowel, and this not only in loanwords, but also in native words. This, in turn, is worthy of attention, because according to most of the current models, it should not have been possible to have a fi nal stressed open syllable in fi nal position in New Kingdom and pre-New Kingdom Egyptian. 65 For instance, forms like ϩⲣⲓ hrˈi: = "endive" and ϭⲗⲏ kʲlˈe: = possibly "strengthen" are usually reconstructed as *hvrˈi:jvj and *qvnˈi:/u:jvj respectively (see e.g. Osing 1976, 102, 193), with a never-attested 66 sequence of weak consonants and glides at their end to prevent the stressed open syllable to be in fi nal position.
The forms discussed here, however, show that there must have been a period before the New Kingdom when such limitation was not valid any more and when open syllables with long vowels could indeed emerge in fi nal position as the result of the fall of a fi nal weak consonant.
As a result, this suggests that the same ϩⲣⲓ hrˈi: and ϭⲗⲏ kʲlˈe: may rather be interpreted as the outcome of a diachronic phenomenon in which the fi nal vowel got lengthened after the disappearance of the fi nal weak consonant, rather than being long before it, because of the presence of an additional syllable. These developments, therefore, can be reconstructed as follow: *hvrˈij > *hvrˈi > *hvrˈi: > ϩⲣⲓ hrˈi: and *qvnˈi/uj > *qvnˈi/u > *qvnˈi:/u: > ϭⲗⲏ kʲlˈe:.
The postulation of such a lengthening of the fi nal stressed vowel after the disappearance of a fi nal weak consonant provides an elegant alternative to the current glide-rich models, at least in the case of those forms displaying a fi nal long vowel in open syllable. §12.

The group -Attestations in the Onomasticon of Amenope
Here below all the attestations of the fi nal group in the Onomasticon of Amenope are presented and singularly discussed. 64 I would say either by being dropped or by being assimilated to the preceding vowel. 65 See e.g. Loprieno 1995, 37, 40. 66 And rather clumsy, I would say. See also HoSW 132-3.173. This word is usually interpreted as a loanword from Semitic (Hoch 1994, 132-4, no. 173) and it is compared with Akk. murʔu = "some kind of offi cial" and Ugr. (pl.) mrủ = mur-ú = "member of a group or class", "chief groom (?)", which in turn may be of foreign (Hurrian?) origin. 67 The fi nal stressed vowel of this word must have been /u:/, because this is the only vowel that consistently remains a back vowel, spelled with w = U, in all three periods. This reconstruction is in agreement with the Semitic evidence.
Two of the spellings attested in Period 2 are worthy of attention. The use of in is unique, and somehow surprising. If it is not just a mere mistake of the scribe, it may have been used either to mark a geminated pronunciation of the /r/ or perhaps to explicitly 67 Hoch 1994, 132-3, no. 173 compares also Arb. ʔimruʔu = "man"; Old Arm. m-r-ʔ = "lord", "sir", but the connection is not clear. Both DULAT 571-2 and CAD 10 228-9 consider the Ugaritic and Akkadian words as possible borrowings. 77 §12.2 The group -Attestations in the Onomasticon of Amenope indicate that the following group had to be read rU and not Ur. The fi nal group in the spelling , which I read as mA.rU.
ʔA/0 = mArUʔA/0, may suggest the presence of a pronounced glottal stop (or glide?) after the fi nal back vowel. This would agree with the Semitic prototype of this word, which indeed had a /ʔ/ as its third and last consonant. The word is attested only in the Onomasticon. It is usually interpreted as a Semitic loanword, related with Akk. sg.
mākiru -pl. makirū = "trader(s)", Heb. sg. moḵer -pl. moḵrīm = "vendor(s)", as well as Punic mkr = "trader", Old. South Arb. mkr(m) = "trader(s)" and Ugaritic mkr(m) = "trader". The Egyptian form seems to be plural, as indicated by the classifi ers . Its vocalisation matches that of the plural of the Akkadian form, and the lack of any trace of a North-West-Semitic -m/-n plural marker supports this interpretation. This observation is interesting, from a socio-linguistic point of view, and may refl ect the status of lingua franca that Akkadian enjoyed in the context of diplomatic and economic exchanges in the region during the Late Bronze Age. The word is attested only in the Onomasticon. Hoch 1994, 328, no. 474 compares it with Heb. kɛlɛʔ = "prison", "confi nement" and with the related root √k-l-ʔ attested in Akkadian, Arabic, Aramaic, Ge'ez, Syriac and Ugaritic with the meaning of "to guard", "to restrain". The vocalisation of the Egyptian word does not match any attested Semitic form. The presence in Egyptian of a stressed fi nal back vowel 78 §12 The group may suggest that the Semitic prototype had a vocalic pattern cvcāc > cvcōc or cvcūc, and the long vowel in the second syllable may have been perceived as stressed by the Egyptians. cvcāc > cvcōc would be a likely candidate, as such patterns are common in Semitic languages. If so, the most likely interpretation is that the Egyptian form derives from a nomen actionis, possibly originally used in an expression comparable to Hebrew bēt kɛlɛʔ = "house of confi nement" = "prison" (see Hoch 1994, 328).
If instead the vowel was ū, then it may refl ect a noun of a dwelling place built on the originally collective pattern cucūc, as it is the case for Hebrew zəḇūl = "habitation", "temple". The use of plural forms for single dwelling such as "house" and "palace" is indeed attested in both Hebrew and Ugaritic (Fox 2003, 209-12 and especially 211-2). The form attested in the Onomasticon is probably a plural, given the classifi ers .
In the past, the word has been connected with Coptic ⲙⲁⲧⲟⲓ = "soldier(s)" (see e.g. the entry DZA 24.499.380 in the TLA). This connection, however, was rejected by Sethe (1916, 124-31), who rather derived the Coptic word from Egyptian m.dU.y = "Medes". 68 This interpretation was accepted by later scholars such as Gardiner (1947, Vol. I, 81*), Černý (1976, 93), and Vycichl (1984, 125). Sethe's intuition can now be further supported on the basis of the vocalisation underlying the word . In fact, as appears from the attestations above, the plural of this word was characterised by a back vowel since Period 1, while a back vowel is attested in the singular at least since Period 2. This means that the word was probably vocalised with a stressed /u:/. Since /u:/ usually develops into ⲏ in Coptic, and not into ⲟ, this Egyptian word can hardly be related with the Coptic ⲙⲁⲧⲟⲓ.

§12 The group
In Middle Egyptian, the word was consistently written with a fi nal -Ꜣ, which at the time was consonantal and which may be related with the of some of the spellings listed here. This may suggest that a glide, or possibly an extra syllable, was present at least until Period 2. Such extra syllable would justify the length of the vowel /u:/. Gloss. Gol. 3.10 (AEO I no. 198) According to Hoch (1994, 268-9, no. 381), this word may be related with the Semitic root √ś-k-y = "to look", "to watch", although no direct prototype or precise parallel can be identifi ed. As discussed elsewhere (Kilani 2017a, 201 "gypsum" (Hoch 1994, 307-8, no. 442), which is usually written without a fi nal back vowel. The fi nal stressed back vowel was likely /o:/ 81 §12.2 The group -Attestations in the Onomasticon of Amenope or /u:/, because of the spellings with back vowel in Period 2. The word = qAḏA "gypsum" is likely a Semitic loanword related with Akk. gaṣṣu, Arb. jiṣṣ, Arm. geṣā, Heb. geṣ, all meaning "plaster", "gypsum". By contrast, Hoch thinks that the word ˹ ˼ = "plasterer" may be an Egyptian formation derived from such Semitic loanword. Gloss. Gol. 3.1,I 555) All the attested forms seem to be plural. appears twice in the Onomasticon, once as an independent entry (no. 555) and once as part of the expression = "maker of sweetmeats" (no. 149). The word is attested also in off ering lists from Medinet Habu, where it is spelled without any back vowel. This may suggest, as implied in the reconstruction off ered here, that the back vowel emerged only in Period 3, as a result of the /a/ > /o/ shift. In fact, if the vowel were /u:/ or /o:/, it should have been marked with U already in Period 2. This is however just a tentative suggestion and should be taken with due caution, both because the Period 2 attestations are not spelled in full group writing and because they come from monumental inscriptions, and therefore they may refl ect diff erent orthographic conventions.  This spelling is a hapax. However, the word can be recognised as an ethnonym, because it appears in the Onomasticon among other names of populations. It has been suggested (see Gardiner 1947, Vol. I, 124*-7*)  The use of the group is clearly optional in words spelled in traditional orthography. This means that its presence indicates the presence of back vowel, while its absence does not necessarily imply the presence of a non-back vowel. Rather, the absence of the group may simply be due to a more conservative and traditional spelling in which the vocalisation was not marked at all. In determining the nature of the vowel lying behind the U, therefore, it is only the presence of the group in a given period that is signifi cant and should be considered, while its absence from some of the attestations is not necessarily meaningful.  4.11 (AEO I no. 306) Cpt.: CrCD 66b; Č eCED 41; VyDELC 53. The Coptic word ⲏⲓ means "pair", "couple" and derives from the Egyptian = ʕ "arm", "hand", as well as "pair", "couple" Vycichl 1984, 53. This spelling, which occurs in the title ẖry-ʕ = "subordinate", "assistant" (lit. "one who is under the hand(s)"), is attested only here. The word itself is obviously extremely common, but it is usually written with the classical spellings or . The use of the group suggests a fi nal stressed back vowel, which can be reconstructed as /u:/ on the basis of Coptic.

Words with Coptic refl exes
The grammatical number of all these expressions is diffi cult to ascertain: although they seem to be considered (in Coptic) and spelled (in Egyptian) as singular, they may ultimately derive from a dual form ʕ.wy. If so, the title should be understood as "one who is under the (two) hands", and the fi nal glide and long vowel could be explained as traces of the original following extra syllable, i.e.*ʕˈu:j < *ʕˈu:wvy. If so, however, one may expect the fi nal glide to emerge in Coptic as -w = -ⲟⲩ rather than as -j = -ⲓ, as in the case of ⲥⲛⲁⲩ snˈaw < *svnˈawwvy = "two" or ⲡⲁϩⲩ pˈahw < pˈaḥwvy = "buttocks". It has to be noted, however, that the vocalic patterns of these two words are diff erent (i.e. (cv)cˈvc+wvy versus cv:+wvy), which perhaps may justify a diff erent development, in the case of , for the fi nal glide derived from the ending -wvy. The Coptic form allows to reconstruct the vocalisation of this word in the three periods as *mˈa:ʾ -*mˈo:ʾ -*mˈo:ʾ respectively, where ʾ stands for either a glide -j or a glottal stop -ʔ developing into a glide. The attested spellings agree with these reconstructions, and support the conclusions about that the group suggested above. By contrast, reconciling these forms with earlier forms is less straightforward. In Middle Egyptian the word was usually spelled , which suggests a disyllabic word *mˈv:Ꜣvj. If this is correct, one may suggest a development *mˈv:Ꜣvj > *mˈv:Ꜣ > *mˈv:j, that is characterised by the fall of the fi nal syllable and by the consequent shift -Ꜣ > -j. This is the interpretation followed for instance by Vycichl (1984, 109). Alternatively, a metathesis may have taken place, thus resulting in the development *mˈv:Ꜣvj > *mˈv:jvꜢ > *mˈv:j. The spellings of all the three periods consistently indicate the presence of a stressed back vowel, which can therefore be reconstructed as /u:/. This word can thus be vocalised as *mvjnˈu:w(ə).

§12 The group
Two main forms are attested in Coptic. The fi rst and most common one is ⲙⲁⲛⲉ = mˈanə < *mˈi/ujnv or the like. 69 The second one, ⲙⲁⲛⲏⲩ (S), ⲙⲉⲛⲉⲩ (B), is rarer. A plural ⲙⲁⲛⲏⲩ (S), ⲙⲁⲛⲏⲟⲩ (B) is also attested. The singular form ⲙⲁⲛⲏⲩ (S), ⲙⲉⲛⲉⲩ (B) has often been interpreted as a compound noun from mjnʾ-ʕw.t = "herdsman of cattle" (Crum 173b implicitly, Černý 84). Vycichl (1984, 115-6), however, disagreed with such interpretation, arguing that ʕw.t should have resulted in -ⲏⲩⲉ, not -ⲏⲩ; he thus wondered if this form could derive from a plural used as singular/collective. I think that Coptic ⲙⲁⲛⲏⲩ (S) is the form underlying the Egyptian spellings, as its New Kingdom vocalisation can be reconstructed as *mvjnˈu:w(ə). As for its origin, I wonder if it is an independent alternative form, as implicitly suggested by Vycichl, or if it is rather the result of a (pre-New-Kingdom) lexicalisation of the expression mjnʾ-ʕw.t, which resulted in a new, etymologically obscure form. I fi nd the fi rst solution somehow easier, but I do not think the evidence available is conclusive. At any rate, the form ⲙⲁⲛⲉ = mˈanə < *mˈi/ujnv could be a secondary form or an innovation that become common only later.
The Period 1 plural , which displays a , may suggest a slightly diff erent vocalic pattern, perhaps something like *mv (jv)nwˈu:y/ w(ə). This form may be at the origin of the Coptic plural ⲙⲁⲛⲏⲩ (S), ⲙⲁⲛⲏⲟⲩ (B), which would then have a diff erent origin from the singular ⲙⲁⲛⲏⲩ (S), ⲙⲉⲛⲉⲩ (B). The diff erence in the Bohairic forms may support this interpretation. This suggestion is however very speculative, as the evidence is not clear enough to be conclusive. Vycichl (1984, 116) mentioned also a personal name ᵐma-ni-e from cuneiform sources contemporary to the 18th Dynasty, which he analysed as a possible variant of this word with the stress on the last syllable. Such a vocalisation, however, matches neither the Egyptian nor the Coptic attested forms, and I actually doubt it is related at all. In the Onomasticon, this word appears in the expression [ ] = "low-lying shoal" (lit. "it comes in the return of the year (?)"see Gardiner 1947, 12*, no. 57).
The spelling seems to be unique and attested only in the Onomasticon. Usually, this word is written according to its classical spelling and the like.
As discussed above, Coptic /e:/ and /e/ can derive from /i:/ or /u:/ and from /i/ or /u/ respectively. On the basis of Coptic, Vycichl (1984, 141) reconstructed the vocalisation and development of a feminine variant of this word as *nˈiryat > *nˈiꜢyat > *nˈiyꜢat > *nˈiy > Coptic nˈej, thus implying a parallel development *nˈi:rvy > *nˈi:Ꜣvy > *nˈi:yvꜢ > *nˈi:y > Coptic nˈe:j for the parallel form with long vowel. The spelling in the Onomasticon, however, suggests that the stressed vowel was a back vowel, not a non-back one. Therefore, I would rather reconstruct the following development: *nˈu:rvy > *nˈu:Ꜣvy > *nˈu:yvꜢ > *nˈu:y > Coptic nˈe:j. In that case, a vowel /a:/ > /o:/ would be a better candidate. However, in words written in standard orthography, the presence of a fi nal is an explicit indicator of the presence of a back vowel, while its absence is not necessarily an indicator of the absence of a back vowel. The current available evidence, therefore, is not conclusive. [ ? n-b-y *nvbˈo:/u:y(ə) [.] n-bUy n-bUy  Gloss. Gol. 5.13 (AEO I no. 424) The word appears in the Onomasticon in the expression = "living room" (lit. "place of sitting"/"place to sit").
The same expression is attested also in Period 1 and 2.
90 §12 The group is likely an infi nitive of the IV-infi rmae verb ḥmsj = "to sit", which in Middle Egyptian would have appeared as ḥmst, with an ending -t replacing the fi nal weak consonant (Allen 2010, 164). By the New Kingdom, however, this fi nal -t was not pronounced anymore, being lenited to -ʔ or possibly completely dropped (Loprieno 1995, 38). Traces of it could survive in writing, as historical spellings, as it is probably the case for of Period 1.
This observation is important, because it means that the fi nal syllable of this word was etymologically closed, being -cvt > -cvʔ, and therefore its vowel etymologically short. This means that the fi nal back vowel represented by in Period 3 in the Onomasticon must have been /o/, because this is the only short back vowel available in the Egyptian vocalic inventory of the time. The attestations from Period 1 and 2, both written without any indication of a back vowel, agree with this interpretation. The vocalisation of the infi nitive of this and comparable IV-inf. verbs, therefore, can be reconstructed as *cvccˈat > *cvccˈoʾ (Period 3), which is likely a variant of the *cvccˈaj/w > *cvccˈoʾ pattern already recognised by Osing (1976, 47-8)  Gloss. Gol. 1.6 (AEO I no. 10); P.Hood I 2.5 (AEO II no. 10) The word is known in all the three periods, and an additional later attestation spelled can be found in Pap. BM EA 10474 (Teachings of Amenemope), dating to the 26th Dynasty (Laisney 2007, 6). This last form points to a vowel /o:/, because as discussed above ( §3.1), by that time the vowel /u:/ seems to have already turned into a non-back vowel /e:/ or the like. The Period 1 spelling (KRI II 173.14), without back vowel, agrees with this interpretation.
In previous periods, this title is usually spelled and the like, which obviously does not provide any information about it vocalisation.
See here below for discussion. 92 §12 The group §12.3 Words in construct state -Some observations = "commander" and = "builder" have direct descendants in Coptic, namely ϫⲟ(ⲓ)ⲥ and ⲉⲕⲱⲧ. However, the syntactical state of the Egyptian forms underlying these spellings is diff erent from that attested for their Coptic descendants, the fi rst being in the construct state and the latter being attested only in the absolute state. The diff erence between these two states is essentially prosodic: on the one hand words in the absolute state are either autonomous prosodic units or core elements of larger prosodic sequences, while on the other hand words in construct state are prosodically subordinate elements within larger prosodic units, which usually consist of simple direct genitival constructions or direct genitival constructions lexicalised as compound nouns. The expressions , "maker of little (vessels?)" (no. 184), = "maker of hn-vessels" (no. 185), ˹ ˺ ˹ ˺ = "builder of walls" (no. 186), and = "commander of the archers" (nos 234-5) discussed here (see below) may belong to the latter case, both because they are listed as entries in a lexical list, which may suggest they were perceived as single lexical units by the Egyptians, and because simple direct genitival constructions are relatively rare in Late Egyptian.
The prosodic profi le of a word in absolute state depends exclusively on morphological features (e.g. number or gender) of the word itself, while the prosodic profi le of a word in construct state may and do change due to its subordinate relation with other prosodic elements of the sentence. Simple direct genitival constructions and compound nouns, and therefore the construct state, were common features in Middle Egyptian, but they are rare and restricted to a limited number of words in Coptic. The construct state is therefore attested in Coptic only for a few words, and it is generally characterised by the absence of any stressed syllable and therefore by the reduction of all its vowels (e.g. Cpt. ⲙⲛⲧⲣⲱⲙⲉ = məntrˈo:mə < *mˈanvt + *rˈa:mvt -see Loprieno 1995, 57). Moreover, scanty evidence from cuneiform transcriptions seems to suggest that the construct state prosodic patterns of simple genitival constructions and of compound names may have been diff erent, in ways that are far from being completely understood (Loprieno 1995, 56-7). Therefore, it is generally impossible to use Coptic forms in their absolute states to reconstruct the vocalisation and stress patterns of construct states in previous periods.
For these reasons, it is not possible to compare the forms and , in construct state, with their Coptic descendant, attested only in absolute state. What is possible to do, instead, is to use the group writing spelling of these words to infer some information about the vocalisation of the construct state of these words in Period 3.
On the basis of the Coptic forms, the vocalisation of the absolute state in Period 3 can be reconstructed as ϫⲟ(ⲓ)ⲥ < *ṯˈoys(ə) and ⲉⲕⲱⲧ < *ʔvqˈo:də, which refl ect the earlier vocalic pattern cˈoccv(y/w) < cˈaccv(y/w) and cvcˈo:cv(w/y) < cvcˈa:cv(w/y) respectively. The group writing spelling of these words can instead be interpreted as *ṯv(yv)sˈUʾ and *ʔvq(v)dˈUʾ. It thus appears that in contrast with the absolute state, in the construct state the last syllable was stressed and characterised by a back vowel.
In order to determine the precise nature of such stressed back vowels, two considerations have to be taken into account. As mentioned above, according to the standard models, fi nal §12.3 Words in construct state -Some observations stressed syllables in native words should be closed, and therefore their vowels should be short. One may wonder if such rule, if it existed at all (see above), was relevant also for the fi rst elements of compound nouns, because in those contexts, those syllables would not be prosodically fi nal any more. In such cases, it may have been possible, at least in theory, that an originally unstressed fi nal open syllable 71 became stressed in the construct state, and could manifest itself as stressed open syllable, because although it was still the fi nal syllable of the fi rst morpheme, it was not prosodically fi nal any more.
This considered, three diff erent vocalisations can be suggested. If their stressed fi nal syllables were closed and their corresponding vowel were short, then these words must have been vocalised *ṯv(yv)sˈoʾ and *ʔvq(v)dˈoʾ, because /o/ was the only short back vowel available in Period 3.
By contrast, if the fi nal syllable was open, then these U may have stood for either a vowel /o:/ or a vowel /u:/, and therefore may have been vocalised as *ṯv(yv)sˈo: or *ṯv(yv)sˈu:, and as *ʔvq(v)dˈo: or *ʔvq(v)dˈu:. On the basis of these words, it can thus be tentatively suggested that the construct state of the vocalic patterns cˈoccv(y/w) < cˈaccv(y/w) and cvcˈo:cv(w/y) < cvcˈa:cv(w/y) was characterised by the movement of the stress to the last syllable and therefore had the form cvc(v)cˈoʾ ~ cvc(v)cˈo: ~ cvc(v)cˈu: < cvc(v)cˈaʾ ~ cvc(v)cˈa: ~ cvc(v)cˈu: or the like. Further evidence is needed to determine whether this behaviour is specifi c of these two words, or it is rather characteristic of these (and other?) vocalic patterns in general. 71 Whose possible existence has been variously suggested -see e.g. Loprieno 1995, 36, 62-3. §13 The case of ym -ⲉⲓⲟⲙ As it appears from the previous paragraphs and from Appendix A here below, the spelling of the word ym -ⲉⲓⲟⲙ appears to be generally regular, with a stressed non-back vowel for Periods 1 and 2 (when the word expected vocalisation was *yam -see Appenidx I.1, II.2) and with a back vowel in Period 3 (when the word expected vocalisation was *yom -see III.2).
There are, however, a few additional attestations which appear to be related with this word and which seem to be somehow irregular. Two in particular need attention.

1) -Astarte Papyrus
The fi rst of such attestations, spelled , comes from the Papyrus Astarte (pBN 202 1.2, 1.x+6, 1.x+13, 2.x+2, 2.x+6, 2.x+11, 2.x+18, 17.y -see Collombert and Coulon 2000), dating to the reign of Amenhotep II. Since the text comes from Period 1, if the word was the same as ym = ⲉⲓⲟⲙ, one would expect a spelling implying a non-back vowel. However, the form clearly suggest a back vowel. This becomes even more problematic if we considered that at the time, not only the shift /a/ > /o/, but also the shift /a:/ > /o:/ had not occurred yet.
The fi rst aspect to consider is the nature of the text (see Collombert and Coulon 2000 for discussion). The papyrus of Astarte is an Egyptian text which records a mythical tale of evident North-West-Semitic origins related with the storm good Ba'al, identifi ed in the text with Seth. The story may have had various episodes, as it was the case for the Ugaritic cycle of Ba'al, but only very small fragments of the papyrus survive, and only an episode of a fi ght between the storm god Ba'al/Seth and the sea god Yam can be identifi ed. The language and style of the papyrus is also worthy of attention: although it is written in proper Late Egyptian, the text contains multiple unusual expressions that have perfect parallels in North-West Semitic sources. 72 With this in mind, the fi rst thing to note is that the word of P.Astarte does not refer just to the "sea" as an extension of water, in general, but rather it refers to the sea as a divine entity. Such deity clearly corresponds to Yammu, the Cananite god of the sea, whose name is indeed identical with the word for "sea" in North-West-Semitic languages. This distinction in the meaning of the word is crucial for two reasons. First, being this word a personal name of a specifi c god, it may have to be considered as lexically distinct from the word ym -ⲉⲓⲟⲙ. Second, this Egytian form could be based on, and could thus refl ect, the pronunciation of this divine name in a specifi c North-West-Semitic tradition, rather than being a transcription of the contemporary pronunciation of the general noun ym -ⲉⲓⲟⲙ = "sea". This possibility becomes even more likely if the narrative preserved in the papyrus was based on a North-West-Semitic original. 96 §13 The case of ym -ⲉⲓⲟⲙ 2) -P.Anastasi I The second exceptional attestation is spelled and appears in P.Anastasi I 21.1, which dates to Period 2. In this case, judging from the context, the word does seem to mean just "sea": "They tell of another city in the sea, Tyre-the-port is its name." The spelling clearly suggests the presence of a back vowel, but in Period 2 a non-back vowel would be expected. Once again, the fi rst aspect to consider is the nature of the text in which this attestation appears. P.Anastasi I preserves a copy of the so-called Letter of Hori, a text relatively widespread in the New Kingdom and likely used for didactic purposes (see Gardiner 1911, passim;Fischer-Elfert 1992, passim, Allen in Hallo and Younger 1997-2002. One of the characteristic features of this text is the abundance of North-West-Semitic words. 73 Moreover, it is also worth noticing that the word appears in a passage that specifi cally describes the city of Tyre, which is located in Lebanon and was one of the main Phoenician centres during the Iron Age.
Besides the two attestations just discussed, a few occasional comparable forms spelled with a back vowel exist also in other documents. 74 Such forms may be all related, and they may share the same origin as those of P.Astarte or P.Anastasi I, or may have been infl uenced by comparable forms.
In this case, one may suggest that the word of the Astarte papyrus had to be read as *yum(mv)~*yu:mV in Egyptian, 75 while in the case of Pap. Anastasi I a reading *yo:mV could also be possible. This form may refl ect a North-West-Semitic prototype *yum(mu) ~ *yūm(mu) ~ *yōm(mu) which, however, does not seem to be attested: the available North-West-Semitic evidence seems to point to a contemporary prototype *yam(mu), with a non-back vowel /a/, not with a back vowel /u(:)/~/o:/. 76 These exceptional spelling are, therefore, diffi cult to explain. If they are not mere mistakes of the Egyptian scribes, these Egyptian spellings may refl ect an otherwise unattested secondary North-West-Semitic post-Cananite shift form *yōm(mu) ~ *yōm(u), deriving from a pre-Cananite shift prototype **yām(mu) ~ **yām(u), with a long vowel. If this is the case, a main Egyptian form *yam(mV) < N-W-Sem. *yam(mu) and a secondary, les common 73 Including a whole North-West-Semitic sentence appearing in P.An. I 23.5. It is possible that one of the purposes of this text was to familiarise the Egyptian students to words in foreign languages that may have been useful in their future career as scribes, hence the high concentration of North-West-Semitic words and loanwords. 74 In particular P.Harris I and P.Turin 21 and 22 (Pleyte and Rossi 1869) -see digitalised slips for ym in the online database of the Berlin Wörterbuch der ägyptischen Sprache. 75 With U = /u(:)/ as at the time this was the only back vowel available in the language. 76 The Phoenician form may have been been /yom/, but the Phoenician development of short /a/ into /o/ is usually considered to be later than the Egyptian texts discussed here, and therefore can hardly explain these spellings. 97 §13 The case of ym -ⲉⲓⲟⲙ form *yum(mu) ~ *yūm(mu) ~ *yōm(mu) may have coexisted at least until Period 3, when the vocalic shift /a/ > /o/ made them indistinguishable in writing.
It is clear that the interpretative model presented here still needs to be refi ned. However, it is the fi rst system that allows to explain all the forms of a methodologically sound and solid corpus, and does so by attributing only one vocalic value to each group -although it does allow for two possible readings CV/VC. The validity of the system can be argued both on synchronic and diachronic ground, and it is strongly supported by the results of the statistical analysis described in §9.
Some aspects, however, still need to be clarifi ed and, hopefully, explained. For instance, it is still unclear if monoliteral signs were functionally diff erent from c+Ꜣ groups. The evidence analysed here seems to suggest that they were equivalent and no diff erence existed between them. However, further studies could help to support this conclusion, or could lead to alternative interpretations.
The same stands true also for the question of the sign . I think that the evidence presented in this book convincingly shows that the sign was not used as a vocalic marker and illustrates some of its other specifi c functions. Nevertheless, in some contexts the reasons for the presence of still remain unclear, and would need further investigation. A systematic reanalysis of Semitic loans attested in group writing, whether they survive in Coptic or not, may help in clarifying these aspects.
The interpretative model presented here, and in general a more reliable understanding of the functioning of the vocalisation of group writing, opens numerous doors for further research. First and foremost, this new interpretation provides an innovative powerful tool to explore the vocalisation of the Egyptian language and its evolution through the Egyptian texts themselves, thus providing a considerable amount of fresh data. Its potential is not only limited to forms attested in Coptic, as those forming the corpus used in this study, but it extends to any word written in group writing, even to those known only from Late Egyptian sources. This is particularly true for those terms that are attested over more than one period: by combining the data from the diff erent attestations, it would be possible to guess not only if the word had a back or non-back vowel, but also the specifi c nature of such vowel. For instance, if a word is attested with a non-back vowel in Period 1 and with a back vowel in Period 2, we can assume that such vowel was an /a:/ that turned into an /o:/, as this is the only vocalic change that took place at that time that could explain such a diff erence in spelling. By contrast, for instance, a word displaying a non-back vowel in Period 2 and a back vowel in Period 3 would imply the presence of an /a/ turning into /o/, while a word displaying a back vowel in Period 1, 2 and 3 would imply the presence of a vowel /u:/ or /u/, as those are the only back vowels that remain stable across the three periods.
The possibility of recognising the nature of the stressed vowel would also provide information about the syllabic structure of such word, as the presence of an /a:/ > /o:/ in Period 1-2 would imply the presence of an open syllable, and therefore of a specifi c syllabic structure for the whole word. By contrast, the presence of an /a/ > /o/ would 100 §14 Conclusions imply the presence of a closed syllable in Period 1-2, and therefore an altogether diff erent syllabic pattern. 77 More in general, to be able of reconstructing the vocalisation also for words which are not attested in Coptic or in other external sources 78 provides a whole new range of possibilities to deepen our understanding of various linguistic aspects of the Egyptian language. For instance, the vocalisation of verbal forms could be studied. For the reasons explained in §3.1, no verb was included in the corpus used in this book. Nevertheless, various verbs are indeed attested in group writing, and their study could shed new light on crucial grammatical issues.
The vocalisation inferred from group writing spellings could also be used to date the texts in which they appear. It is clear that if the group writing spelling of a word in a given text suggests a stressed /o:/, while that of another word in the same text suggests a stressed vowel /a/, then the that text should likely be dated 79 to Period 2, because only then the vowels /o:/ and /a/ could coexist.
Similarly, the appearance of Semitic loanwords attested in group writing could be used to estimate when they have entered Egyptian, for instance by correlating their vocalisation with the so-called Canaanite vocalic shift. This in turn may help in better defi ning the chronological and sociocultural frames of the interactions that brought these words into Egyptian.
More in general, this new reading of group writing words could certainly bring new crucial data also for the study of (North-West-)Semitic languages and dialects, for a period for which only few scanty traces are otherwise available.
Finally, this new interpretation of group writing could also be applied to historical sources, such as the topographical lists, not only to obtain a better reading of the names themselves, but also to verify the identifi cations suggested so far, which until now have usually been based only on the consonantal skeleton of these names. 77 This, for instance, could be used to dismiss doubtful Coptic etymology. For instance Černý (1976, 340) suggested to link Coptic ϭⲟⲩϫ "saffl ower" "cardamom" with Egyptian kṯ "some herb or fl ower". This etymology was however very doubtful and was not endorsed by any other scholar. Now it can also be rejected on the bases of the Egyptian evidence itself. The Egyptian word kṯ is in fact spelled = kAṯ(A) in Period 1, = kAṯ(A) in Period 2, and = kUṯ(A) in Period 3. This sequence A-A-U clearly indicates the presence of a stressed short /a/ shifting to /o/. The Coptic form, however, requires a long /a:/, which would have shifted to /o:/ already in Period 2. Therefore, the Egyptian and Coptic forms cannot be directly related. 78 Such as Akkadian transcriptions. 79 Either in its composition or in its redaction -this is actually an aspect that indeed need to be further researched. Syr. ʔaggānā, all referring to some kind of vessel. Notes: the form a-ku-nu appearing in an Amarna letter is likely a transcription of the Egyptian form, rather than of Semitic prototype, both because of the k instead of g and because the word appears in a list of vessels sent from Egypt. If so, the spelling ku would confi rm once again the early shift /a(:)/ > /o:/ in Egyptian (see above §4.5.1). It is worth noticing that a variant is also attested (LeLE i.50 H.O. 87v5). This form is clearly spelled according to the w-orthography and has to be read as jkn-w = jkwn. This spelling is important because it confi rms the back nature of the stressed vowel and therefore confi rms the reading of the form spelled in group writing. Finally, two variants ⲉⲕⲱⲛⲉ and ⲉϭⲱⲛⲉ are also attested in Sahidic.  Cpt.: CrCD 3b; ČeCED 4; VyDELC 6; WeKH 3 Notes: Possibly plural or collective. No diff erent form for the plural is attested in Coptic, and it is possible that this was the case also in Late Egyptian. Alternatively, it is likely that the plural had a structure *ʕˈi:/u:lv. In this case, the length of the stressed vowel would have been diff erent, while its nature would have been the same as in the singular. Cpt.: CrCD 16b; ČeCED 12; VyDELC 16; WeKH 12 Sem.: *ʕad(a)ši:n (pl.) -cf. Heb. ʔă dāšīm (pl.) and see Arb. ʔadas (col.), both "lentils"; note that the Egyptian form derives from a Semitic plural with nunation, as it could be expected from Aramaic, rather than from a plural with mimation like in Hebrew. Moreover, the Coptic form suggests the syncope of the middle vowel, although it is impossible to say if such syncope occurred in Egyptian or characterised already the Semitic prototype, and in this respect a Semitic form *ʕadšīn (pl.), however, would certainly not be surprising or problematic. Eg.: LES 2 8.4,8.9,10.3,12.6,17.4 = LeLE i.328;P.Harris I 5.2,8.4,8.12,21a.2,21a.6,21a.7,21a.8,21a.9,21a.10,21b.1,29.6,36b.11,40b.3,47.11,49.8,56a.8,73.5  Cpt.: CrCD 765b; ČeCED 312; VyDELC 325; WeKH 418 Sem.: no clear Semitic etymology can be proposed, but according to Hoch the word may be connected with Sem. √zll "shake", from which Heb. zalzallīm "twigs", "branches or tendrils of grape vine". The required semantic shift, however, suggests caution. Notes: Possibly plural. No diff erent form for the plural is attested in Coptic, and it is possible that this was the case also in Late Egyptian. Alternatively, it is likely that the plural had a structure *ḏˈi:/u:lv. In this case, the length of the vowel would have change in respect to the singular, but not its nature. The Egyptian word is often attested in a reduplicated form. Such forms, however, does not match the words attested in Coptic, and therefore they are not discussed here.  App. B Fig. 6: probabilities of random matches for trisyllabic words in Period 2.