2005 Features of Central Semitic more

From Biblical and Oriental Essays in Memory of William L. Moran, ed. Agustinus Gianto. Biblica et Orientalia 48. Rome: Pontificio Istituto Biblico, 2005. Pages 155-203. Features of Central Semitic*, JOHN HUEHNERGARD Harvard University The present study1 is dedicated to the memory of William L. Moran, whose work continues to illuminate and to inspire. It is thanks to Bill Moran's pioneering work on the syntax of the Canaanizing Amarna texts that the verbal system of early Canaanite is known to have been so simi- lar to those of Ugaritic and Arabic, systems that may well reflect that of their common Central Semitic ancestor.2 I. INTRODUCTION The internal subgrouping of the Semitic language family has long been a matter of debate. The prevailing view for much of the twentieth century was that of such scholars as Hommel, Noldeke, and Brockel- mann.3 In this view there is a three-part division (see Figure 1): Ak- kadian is the sole member of East or Northeast Semitic; Hebrew, the other Canaanite languages, and Aramaic comprise Northwest Semitic; and Arabic, Ethiopian Semitic, the Old South Arabian languages, and the * Earlier versions of this paper were read at the Institute for Advanced Studies of the Hebrew University in April 2002 and at the Hebrew University Department of Linguistics seminar in May 2002. For criticism, comments, and suggestions for improvement, I wish to thank J. Blau, B. Fortson, W. R. Garr, G. Goldenberg, J. Hackett, R. Holmstedt, S. Izre'el, J. Joosten, R. Ratcliffe, and A. Rubin. Responsibility for the opinions expressed herein, of course, rests with me. 1 Note the following abbreviations: C = causative stem; CS = Central Semitic; MSA = Mod- em South Arabian languages (Mehri, Harsusi, Hobyot, Jibbali, Soqotri, etc.); OSA = Old (Epi- graphic) South Arabian languages, or Sayhadic (Sabaean, Minaean, Qatabanian, Hadramitic); PNWS = Proto-Northwest Semitic; PS = Proto-Semitic; WS = West Semitic. 2 Moran's publications on Amarna, including his often-cited but previously unpublished dis- sertation, have now been collected into a single volume entitled Amarna Studies (2003). 3 Hommel 1883,442; Noldeke 1899,15-20; Brockelmann, GVG 1.6; for surveys see Golden- berg 1977,473^78; Voigt 1987b, 1-2; Faber 1997, 5-6. 156 JOHN HUEHNERGARD Modern South Arabian languages comprise South or Southwest Semitic. Figure 1: Brockelmann, etal. (after Faber 1997) Common Semitic West Semitic Northwest Semitic South Semitic East Semitic Canaanite Aramaic Arabic MSA OSA Ethiopian Akkadian In a number of articles and monographs published in the 1970's, Rob- ert Hetzron challenged this view.4 Hetzron articulated two principles that he considered fundamental for genetic classification. One is the principle of Archaic Heterogeneity: "when cognate systems (i.e. paradigms) in re- lated languages are compared, the system that exhibits the most inner heterogeneity is likely to be the closest to the ancestor-system"; the sec- ond is the principle of Shared Morpho-lexical Innovations: "the phonetic shape of morphological items is the least likely to be borrowed (as against lexical items)."5 Hetzron proposed the branching seen in Figure 2.6 First, Proto-West Semitic hived off from the common Semitic ancestor, characterized by a new means of expressing the perfective aspect, the suffix-conjugation, 4 See especially Hetzron 1974; 1976. 5 Hetzron 1976, 89. 6 Hetzron was preceded in this view by Christian 1919-20; see Voigt 1987b, 2. FEATURES OF CENTRAL SEMITIC 157 which in Proto-Semitic7 and in Akkadian was simply a conjugated adjec- tive.8 The speech forms that did not participate in this innovation are la- belled East Semitic, and comprise Eblaite and the various forms of Ak- kadian.9 Proto-West Semitic, in Hetzron's scheme, in turn splits into two branches: South Semitic and Central Semitic. On the basis of earlier studies by Joseph Greenberg (and others), Hetzron plausibly assumed that the imperfective form of the verb had the shape we find in Ak- kadian, Ethiopic, and the Modern South Arabian languages, namely, a form with a two-syllable base and a doubled middle radical, *yVqattVl. Hetzron then suggested that the languages in which this Proto-Semitic form has been abandoned and replaced with a very different form, *yaqtulu, namely, Arabic, Aramaic, and the Canaanite languages, must have shared a common ancestor, an ancestor that he labelled Central Se- mitic. Thus, Proto-Central Semitic, with its innovative imperfective verb, splits off from Proto-West Semitic. The remaining part of West Semitic, which he called South Semitic, consisted for Hetzron of Ethiopian Se- mitic, Old South Arabian, and Modern South Arabian. 7 Possibly also in Proto-Afro-Asiatic, or at least in a common ancestor of Semitic and Egyp- tian. 8 This earlier usage is retained secondarily, inter alia, in Biblical Hebrew in forms such as zoqdnti 'I am old' = 'I have become old'. 9 Akkadian and Eblaite also share a number of innovations, such as the ending - fit for mascu- line plural adjectives, and the development of dative pronouns. Eblaite should probably not be considered a dialect of Akkadian (pace Dombrowski 1988; Krebernik 1996), but rather a co- equal branch of East Semitic, since Eblaite does not participate in a number of pan-Akkadian innovations, such as the change of initial m to n in maprVs nouns from roots containing a labial consonant (as in Eblaite ma\bahum vs. Akkadian nebehum 'belt'; see Edzard 1981, 132), the re-grammaticalization'of f-conjugation preterites as forms of a new tense, the "perfect" iptaras (von Soden 1984, 20-21), and the shift of the verb to clause-final position (Eblaite is often verb- initial; see Edzard 1984); conversely, Akkadian dialects do not exhibit the Eblaite innovation of t-form verbal nouns with both prefixed and infixed t's (Gt taptaris, Dt tuptarris, St tuStapris; see Hecker 1984; Kienast 1984; Krebernik 1984). 158 JOHN HUEHNERGARD Figure 2: Hetzron (1976, etc.) Common Semitic Central Semitic Arabo-Canaanite East Semitic South Arabian Ethiopian Canaanite Arabic Aramaic MSA OSA N. Eth S. Eth. Akkadian As can be seen, it is the position of Arabic that is the crux in these competing models. A number of scholars have rejected Hetzron's model, for a number of reasons. For some, Hetzron's choice of the imperfective form of the verb as a diagnostic innovative feature is unacceptable.10 It is, however, a feature of basic and deep structural importance; the imper- fective form of Hetzron's Central Semitic, the yaqtulu form, is funda- mentally distinct from the yVqattVl of Akkadian, Ethiopic, and Modern South Arabian. If the yaqtulu imperfective is an innovation, it is a sig- nificant shared morpho-lexical innovation; if it is not an innovation, it must have been inherited from Proto-Semitic, that is, it must have been the Proto-Semitic imperfective verb. That, in turn, implies one of two things: (1) either yaqattal in Akkadian, Ethiopic, and Modern South Arabian is a shared innovation in those language groups; or (2) it is an innovation that occurred independently in each of those groups. The lat- ter view must be rejected; the form is so similar in those languages, in- cluding, as shown recently by Gensler (1997), the doubling of the third radical in the imperfective form of quadriradical roots in both Ethiopic and Akkadian, that the notion that each group came up with it, essen- tially ex nihilo, stretches credulity. The former idea, that yVqattVl is a > See, e.g., Blau 1978; Diem 1980; Zaborski 1991; Knudsen 1998. FEATURES OF CENTRAL SEMITIC 159 shared innovation in the groups that exhibit it, also fails to convince, since the perfective suffix-conjugation, qatala, would then have to be assumed to arise in two distinct subgroups, viz., all of the yaqtulu group and part of the yVqattVl group. Finally, the assumption that yVqattVl is not of common Semitic origin would also mean that it must be unre- lated to similar forms attested in other branches of Afro-Asiatic, as shown by Greenberg and others. Other scholars have questioned Hetzron's model because it allows no genetic significance to the broken plurals that characterize Arabic, Old and Modern South Arabian languages, and northern Ethiopian Semitic. Indeed, recent studies by R. Ratcliffe (1998a, 1998b) have pointed out that the patterns of broken plurals exhibited in these language groups suggest a common ancestor. But as others have noted, that ancestor is more likely to be Proto-Semitic, or even Proto-Afro-Asiatic.11 It is widely recognized that there are vestiges of broken plurals in the North- west Semitic languages, and a few relic forms have also now been pos- ited for Akkadian,12 so there can be no doubt that the phenomenon of broken plurals is a feature of Proto-Semitic. It may be suggested that the similarity of the patterns that Ratcliffe has found in Arabic, the Old and Modern South Arabian languages, and northern Ethiopian Semitic is due to an expansion of the broken plural repertoire in the Proto-West Semitic period; these plural formations were then largely abandoned by the Northwest Semitic languages, except for the well-known mslek ~ mlokim type, and a number of relic forms.13 In other words, Proto- Semitic probably had a relatively restricted set of broken, or pattern- replacement, plurals, inherited from Afro-Asiatic. In East Semitic, these were essentially given up, while in West Semitic, the system was not 11 See Goldenberg 1977, esp. 473^175. 12 See Huehnergard 1987d, 181-188. 13 Such as rekeb 'chariotry, riders'; perhaps also forms such as the collective/plural peproah 'young birds'. See also Gordon 1991. Note also perhaps Phoenician Dgdd 'gang of bandits' (but see Garr 1985, 47), aqtl forms in Ugaritic (but see Moscati 1957), and Man "Amorite" forms such as aqdamu 'earlier time', aqdamdtu 'near/eastern bank' (see Streck 2000, 84). 160 JOHN HUEHNERGARD only retained, but further modified and expanded to what we find in most of the descendant branches. Then, in Proto-Northwest Semitic, the sys- tem was largely abandoned, as it had been earlier in East Semitic; it should be noted that the same process can be observed in other descen- dant branches, such as Southern Ethiopian Semitic, where the internal plural system has likewise been lost. Other features shared by Arabic and either Ethiopian Semitic or Mod- ern South Arabian languages, such as the L-stem conjugation qatala and the vowel melody a...a of the suffix-conjugation of the derived stems (qattala, Daqtala), should also be ascribed to an earlier node, probably to the Proto-West Semitic period.14 Finally, the change of *p to f, which is also frequently mentioned as tying Arabic together with Ethiopic and Modern South Arabian,15 is, to quote A. Faber, "so natural as to be useless for subgrouping."16 Thus, Hetzron's basic model remains the most economical and accurate depiction of the genetic subgrouping of the Semitic language family. A number of scholars, while generally accepting Hetzron's classifica- tion model, have proposed modifications to it. One change was to get rid of Hetzron's "Arabo-Canaanite" subbranch when the feature that, for him, tied Arabic and Canaanite together, viz., the ending *-na on femi- nine plural verbs, was also shown to have occurred in early Aramaic; we will return to that feature in more detail below (see §11, 6). Perhaps the most significant modification is a result of an important study of the verb in the Old South Arabian languages by N. Nebes (1994). By examining forms of weak verbs, Nebes demonstrated that none of the languages for which there is sufficient evidence - Sabaean, Minaean, and Qatabanian - exhibits the form yVqattVl; the imperfec- 14 Goldenberg 1977, 475. Thus, *qattila and *haqtila would be innovative in Proto- Northwest-Semitic, against what I concluded in Huehnergard 1992. 15 The phoneme was probably Ifl also in OSA, or at least in Sabaean (see Beeston 1984, 10), which is now also to be grouped with Central Semitic; see further below. 16 Faber 1997,4. FEATURES OF CENTRAL SEMITIC 161 tive form of the verb is, instead, yaqtulu}1 This is very significant: it means that these languages participated in the principle innovation that characterizes the Central Semitic branch, a point to which we will return a number of times in what follows; it also means that none of these lan- guages can be the ancestor of either the Modern South Arabian lan- guages or the Ethiopian Semitic languages. Another modification to Hetzron's model was proposed by V. Pork- homo vsky in a recent short paper (1997), in which he pointed out that Hetzron's South Semitic branch may be an illusion, since it is based on a shared retention, namely, the imperfective yVqattVl form, rather than on any well-defined shared innovations. In other words, it is simply what is left of West Semitic once Central Semitic hives off. Thus, Porkho- movsky suggests that, until some significant innovation shared by Ethio- pian Semitic and the Modern South Arabian languages has been identi- fied, they should not be considered a genetic subgroup, but rather sepa- rate descendant branches of West Semitic. One feature that is shared by Ethiopian Semitic and by Modern South Arabian is the change of t to k in the second person forms of the suffix-conjugation, that is, *qatalka instead of *qatalta}z As will be seen below (§11, 5), however, the sig- nificance of this feature for genetic classification is uncertain. Thus in Figure 3, Ethiopian Semitic and Modern South Arabian are provisionally shown as separate branches. 17 So also Voigt 1987b, 6-7. 18 See Hetzron 1976, 93-94; Faber 1997, 11. Faber, ibid., also suggests another shared inno- vation of Ethiopic and MSA, namely, the generalization of *Dal as a verbal negative. Note, how- ever, that some of the MSA languages, at least, use *la (Simeone-Senelle 1997, 413—414). Fur- ther, OSA Sabaean, which is now to be grouped with Central Semitic, uses Dl (vs. Minaean Ihm). Thus, perhaps we should see this either as an areal feature or as an instance of parallel develop- ment. Rodgers 1991, 1328-1329, considers diagnostic the fact that "Verbal noun forms in Sabaic and Ethiopic evolved toward use as verbs, while they retained their original role in the Modern South Arabian languages." (See also Faber 1997, 11.) These constructions are quite different, however; the Ethiopic forms are "conjugated" with (nominal) suffixed pronouns, whereas the OSA forms seem to be extensions of the use of the infinitive in narrative prose that is is found, for example, in Hebrew, Phoenician, and Ugaritic. 162 JOHN HUEHNERGARD Figure 3: Hetzron's Classification with Modifications Common Semitic West Semitic East Semitic Ugar. Canaanite Aramaic Arabic OSA MSA Ethiopian Eblaite Akkadian The continuing debate about the subgrouping of Semitic is salutory. It is intrinsically of interest to try to understand what the ancestor of, for example, Northwest Semitic was, what it looked like, and what its clos- est affiliations were. But more important, perhaps, is the fact that linguis- tic history provides valuable evidence for the history and even the pre- history of groups of speakers. Comparative linguistics seeks to trace the histories of related languages and to account for genetic relationships, and in order to do so it must make claims about internal relatedness and subgrouping. If, for instance, Northwest Semitic shares a closer common ancestry with Arabic and Old South Arabian, then features that those languages share provide important information about the history and de- velopment of Northwest Semitic. In this paper we will review a number of features that are apparently restricted to the languages of the Central Semitic branch, including fea- tures that have been noted by other scholars, especially Faber in her 1980 dissertation and in a number of important articles.19 The significance of these features will be considered both with regard to whether they are "Faber 1980; 1985; 1990; 1991; 1997. FEATURES OF CENTRAL SEMITIC 163 further indications of genetic subgrouping and with regard to what they add to our picture of Proto-Central Semitic. Although all features that a group of languages exhibits in common tell us something about the history of those languages, not all such fea- tures are significant for genetic classification. According to standard principles of historical linguistics, the only features that are significant for genetic classification, that is, that are indicative of a common ances- tor, are shared innovations. One of the clearest examples in Semitic is the shape of the perfective form of the verb. In Proto-Semitic that was *yaqtul, as in Akkadian iprus; as has already been noted, the "suffix- conjugation" was essentially a conjugated adjective, again as in Ak- kadian and also as in early Egyptian. In the languages other than Ak- kadian and Eblaite, this conjugated adjective has become the main per- fective form of the verb, relegating the earlier *yaqtul to secondary functions; that development must have taken place in a shared common ancestor, a single proto-language, which we label Proto-West Semitic, that passed on the change to its descendants. But shared features may also be the result of other factors, and it is worth reviewing the most important of those before we proceed. First, a shared feature may have been inherited from a still earlier ancestor, and have been lost in other groups of the family. This is called a shared re- tention, and it is generally not considered relevant for subgrouping, since it does not involve a common intermediate ancestor. We would not group together Akkadian, Hebrew, and Arabic on the basis of the fact that they all have a derived verbal N stem, whereas Aramaic, Ethiopic, and the MSA languages do not. The N stem is undoubtedly a Proto- Semitic feature that has been lost independently in the latter languages. See further below, §§11, 7,8. Another source of apparently shared features is spontaneous inde- pendent development20 or parallel development.21 Analogical changes 20 Hetzron 1976, 97. 21 Parallel development has been discussed in some detail by J. Blau (e.g., 1968, 42-43; 1980). 164 JOHN HUEHNERGARD that are obvious and relatively minor, and that could easily take place in several speech communities, are not generally indicative of a common ancestry; many English-speaking children, for example, say brang in- stead of brought, on the analogy of sing : sang; should there be whole speech communities in which brang has become normative, we would not suggest that they necessarily share a common immediate ancestor. See further below, §11,4. Another historical linguistic phenomenon that results in shared fea- tures is the areal or wave-like spreading of features as a result of contact between speakers of different dialects and languages. Herein lies the dis- tinction between tree and wave versions of language relatedness. Family trees represent genetic splits, ancestors, and subgroupings; wave or areal changes account for isoglosses that criss-cross one another and cross the family tree lines, and are of great importance in causing geographically contiguous dialects and languages to become more alike. As shown by W. R. Garr (1985), many of the Canaanite and Aramaic dialects existed in a continuum of contact for long periods of their history, and it is likely that this was also true for early forms of Arabic and Old South Arabian. In such cases, as for instance also in the Germanic languages,22 it is often very difficult to ascertain whether a given shared feature is the result of shared innovation in a common ancestor or the result of areal spreading from language contact. An example is the articulation of the so-called emphatic consonants, discussed below under §11, 2; see also §§II, 5, 13. II. FEATURES We may now turn to the consideration of some of those features that are characteristic of Central Semitic. 1. *yaqtulu Our first feature has already been mentioned above, namely, the re- placement of Proto-Semitic yVqattVl by yaqtulu as the imperfective 22 See, e.g., Nielsen 1989. FEATURES OF CENTRAL SEMITIC 165 verb; as we have already seen, this is one of Hetzron's main diagnostic features. It may, however, be suggested that the Proto-Central Semitic innovation in the verbal system was actually broader and more complex than merely the form of the imperfect verb. As one studies the tense- mood-aspect systems of Arabic, Ugaritic, Amarna Canaanite, and early Hebrew (and perhaps also early Aramaic, in view of the Tel Dan inscrip- tion with its preterite yqtl forms), one is struck by their overall formal and semantic similarity: besides the imperfective yaqtulu, we may men- tion the volitive or subjunctive yaqtula and the emphatic energic forms. Although the details remain unclear, it seems increasingly likely that the entire tense-mood-aspect system of these languages, both formally and semantically, can be reconstructed to a common ancestor.23 Recent stud- ies on the Sabaean tense-mood-aspect system suggest that it, too, reflects this common ancestor.24 2. Pharyngealization of "emphatic" consonants It is now generally accepted that in common Semitic these were ejec- tives, or glottalic pressure sounds, as in Ethiopian Semitic and in the MSA languages. In Arabic, of course, and in modern Aramaic dialects, they are pharyngealized, and Faber has suggested that this is the result of a Proto-Central Semitic change.25 She notes that Hebrew and Aramaic tD forms that show assimilation, such as Hebrew nistaddoq (Gen 44:16), indicate that the early Hebrew emphatics were also pharyngealized,26 23 All of the relevant elements can probably be reconstructed formally to PS: %yaqtulu is probably formally cognate with the Akkadian subordinative form (iprus-u), and the ending *-na in the plural *yaqtuluna may be related to the subordination markers -na in early Old Babylo- nian (for which see Whiting 1987, 13, 43-44) and -ni in Old Akkadian and Old Assyrian; the form *yaqtula is also found sporadically as a subordinative form in early Akkadian (Gelb 1961, 170-171). For -u ~ -na in Ethiopian Semitic, see Hetzron 1977, 90-92; Goldenberg 1977, 478- 481; Appleyard 2002,414-419. 24 See Tropper 1997. 25 Faber 1980,114-170, 232-233; 1990; 1997, 8. 26 But see, e.g., Zemanek 1996, 51. 166 JOHN HUEHNERGARD because pharyngealization tends to spread, whereas glottalization does not. But phonological developments such as this are often the result of areal diffusion, and it seems equally possible that this is the case with the pharyngealization of the emphatics; in the present state of our knowledge about the phonetics of the early Northwest Semitic languages and of the OSA languages, the matter must remain undecided.27 3. *nahnu An uncomplicated feature that nevertheless illustrates some of the problems involved in trying to understand the historical import of shared features is the shape of the independent first person plural pronoun. Be- low are the attested forms in the various languages: Akk. Ethiopic MSA OSA Arabic Hebrew Aram. Bab. nlnu ndhna (d)nha2S unatt'ed nahnu 3andhnu *3anahna29 Ass. nenu rare nahnu10 The MSA forms are difficult to explain; it is possible that a proto-MSA form *nha is quite ancient; more likely, however, it is the result of a Proto-MSA innovation. In any case, we may, I believe, leave those forms aside. The extra syllable at the front of the Hebrew and Aramaic forms is presumably secondary, the result of levelling from the other first and second forms in the paradigm, something that is also seen in later Akkadian 27 Another probable instance of areal spread is the change of PS s to h in morphemes such as the third person pronouns and the causative, discussed by Faber (1980, 220-222) and in a number of studies by R. Voigt (19.87a; 1994; 1995). The fact that this change crosses otherwise well-established genetic divisions - most obviously in its appearance in only one of the OSA languages - indicates that it spread in waves after at least some of the genetic splits had occurred. 28 See Simeone-Senelle 1997, 387, where (a)nha is given for most dialects; nhan also oc- curs in Jibbali, han in Soqotri. 29 Kaufman 1997, 121, notes that ^anahna "may be ascribed to common Aramaic." 30 The rare biblical form nahnu is probably primitive. FEATURES OF CENTRAL SEMITIC 167 dialects.31 As for the final vowel, Hetzron's principle of archaic hetero- geneity suggests that the -u of the Akkadian and Arabic forms is origi- nal.32 Otherwise, however, we have two incompatible sets of forms: Arabic, Hebrew, and Aramaic clearly require us to reconstruct the vowel a between the n and the h, that is, *nahnu; in the Ga'az form, on the other hand, the original vowel must have been other than a; and the Babylonian Akkadian form, finally, requires an original i.33 In other words, we are probably to reconstruct Proto-Semitic *nihnU, with the Central Semitic languages reflecting a change of i to a before the pha- ryngeal h. But did the latter change occur in a common ancestor, in which there was a tendency to prefer a before a syllable-final guttural, a tendency we observe in the Central Semitic languages themselves? Or are the forms in the Central Semitic languages due to independent devel- opments resulting from that same tendency in the descendant languages to prefer a before a guttural? The fact that forms with a appear without exception in those Central Semitic languages for which we have evi- dence may tip the balance in favor of the first alternative. 4. Feminine singular *-at > -a The loss of the final t in the feminine ending *-at on nouns and/or verbs, as in Hebrew malko < *malkat- and kotbp < *katabat, occurs in quite a few Semitic languages; almost exclusively, they are Central Semitic languages.34 But here, of course, the feature cannot be assigned to a proto-language, for a number of reasons: it does not occur in some 31 For Neo-Assyrian and Neo-Babylonian aninu/anlni see GAG §§41i,j; CAD A/2 122-123. 32 That is, the heterogeneity of the final vowels in Arabic nahnu vs. suffixal -na and in Ak- kadian nlnu, nominative -anu vs. genitive -ni, accusative -niati is probably ancient; both lan- guages indicate that -nU is to be reconstructed for the independent pronoun. 33 Pace von Soden, GAG §41g n. 12, who derives nlnu from *nahnu; the latter should yield nenu in both Assyrian and Babylonian. Assyrian nenu may derive from either *nahnu or *nihnu (see GAG §24e). 34 The phenomenon is also found in MSA Soqotri (Blau 1980, 27; Arabic influence?). 168 JOHN HUEHNERGARD of the languages, such as Ugaritic and the OSA languages; it can be traced in the attested history of some of the languages, such as Arabic; and it does not affect the same parts of speech in the various languages (e.g., nouns and verbs in Hebrew; verbs only in Phoenician; nouns only in Aramaic and Arabic). This suggests that the phenomenon is the result of parallel development, as argued by Blau.35 5. Suffix-conjugation first person singular and second person forms Another example that illustrates the possibilities and uncertainties in- volved in assessing the significance of shared features is the consonant of the first and second person endings of the suffix-conjugation. Hetzron used this feature to exemplify his principle of archaic heterogeneity. In Ugaritic, in Hebrew and other Canaanite languages, in Aramaic, and in Arabic, the first person singular and the second person endings all have t, as in Arabic katabtu, katabta, katabti. In Ethiopic and in the MSA languages, they all have k, as in Gs'sz nabarku, nabarka, nabarki. But in Akkadian, the first person singular has k while the second person forms have t, and Hetzron was surely correct to conclude that the Ak- kadian paradigm is the more original, and that the other languages show the effects of paradigmatic levelling. Akk. Ethiopic MSA OSA lcs -ku -ku -k <-K> 2ms -ta -ka -k <-K> Ugar. Hebrew Aram. Arabic <-T> -ti *-tv -tu <-T> -to *-ta -ta But Hetzron also suggested that the paradigmatic levelling of the kit contrast, to t in some languages and to k in others, reflected "two inno- vative groups,"36 that is, that the levellings reflected shared innovations. But as we have already seen, the Old South Arabian languages have now been shown to have shared in the adoption of yaqtulu as the imperfec- tive form of the verb, and therefore are to be included in the Central Se- 35 See Blau 1968,42-43; 1980. "Hetzron 1976,94. FEATURES OF CENTRAL SEMITIC 169 mitic branch. Yet it is now finally also known, from recently published texts, that at least one OSA language had k in the first and second per- 37 sons. And it has long been known that certain Arabic dialects also ex- hibit k.n It is likely that those Arabic dialects have borrowed this from an OSA language. And it is possible that OSA too had k as a result of borrowing, or areal spreading, from an ancestor of one of the MSA dia- lects. But it is also possible that the kit contrast continued beyond Proto-Central Semitic, and that some, or even many, of the levellings to t or to k were independent developments. Paradigmatic levelling is, after all, a common phenomenon, and in this case there are only two ways for the levelling to proceed. Note, for example, that in later Akkadian (Neo- Assyrian), we also find levelling in the suffix-conjugation paradigm, and not to t as might be expected from Aramaic influence, but rather to k (i.e., parsata is occasionally replaced by parsdka, etc.).39 6. Feminine plural ending *-na The ending *-na of the feminine plural forms of the prefix- conjugation, that is, of forms such as Hebrew tiktdbno and Arabic yak- tubna and taktubna, is also found on feminine plurals in Ugaritic40 and in Old Aramaic,41 and probably also in at least some of the Old South Arabian languages 42 But in Akkadian iprusa and Ethiopic yanbara, we find instead a final long -a, and -a- also appears in the ending -an in 37 Ryckmans et al. 1994, 29. 38 See, e.g., Diem 1973, 75-113; Goldenberg 1977,478 and n. 81. 39 GAG §75b-c. 40 See Verreet 1984,317-319. 41 Huehnergard 1987c; Voigt 1987, 6-7. 42 The Sabaean and Qatabanian 3rd fern. pi. suffix-conjugation forms with final -n (see Bee- ston 1984, 14, 64) are reminiscent of Arabic katabna, and are probably from the same source as in the latter, viz., an analogy with a prefix-conjugation formyaktubna. 170 JOHN HUEHNERGARD later Aramaic dialects. It has therefore been argued that the ending *-na of the Central Semitic languages is a shared innovation.44 It is dif- ficult, however, to discover how such an innovation would have come about,45 and it seems more likely that the ending *-na actually goes back to Proto-Semitic;46 it is then a simple matter to suggest that final -a in Akkadian and Ethiopic and in Aramaic -an is the result of an obvious and natural analogy with the suffix-conjugation, viz., *qatvlu : *qatvla :: *yaqtulu : X = *yaqtula. If that is correct, then *-na in the Central Semitic languages is not an innovation in Proto-Central Semitic, but rather a shared retention from Proto-Semitic, and thus not significant for grouping the Central Semitic languages together. A general problem here, it must be noted, is that there is often uncertainty or disagreement, as in the example of -na, as to what is to be reconstructed to earlier lev- els such as Proto-Semitic; that is, there is frequently a lack of consensus as to what Proto-Semitic looked like. The change of *-na to -a in Akkadian, Ethiopic, and Aramaic is an example of independent or parallel development; because of the obvious nature of the analogy that produced it, it is not a feature that should be used to group together the languages that exhibit it. * * * The next three features (7, 8, 9) involve forms of weak verbal roots. A fully triconsonantal root is assumed here for each of these verb types for 43 Also perhaps in the few OSA Sabaean examples that are written qtly; since final -y also appears in Sabaean on dual verbal forms (masc. qtly, fem. qtlty), which presumably ended in /-a/ originally (cf. Arabic qatala, qatalta; Akkadian parsa, parista), the fem. pi. forms qtly probably also reflect an original *qatala. 44 E.g., Hetzron 1976, 103; Tropper 2000, 445 §73.233.43. 45 Hetzron 1976,103, citing Rossler, suggests that the ending *-na is "based on the ending of the corresponding independent pronouns," i.e., hunna, etc.; but it is more likely that -na on hunna, etc., are based on the verbal endings. 46 So also Goldenberg 1977, 477. Voigt 1987b, 11, cavalierly dismisses Goldenberg's argu- ments with the claim that "it is best to posit a paradigm that exhibits only vowel-initial endings." FEATURES OF CENTRAL SEMITIC 171 Proto-Semitic. There have been numerous suggestions that some of these represent original biradical roots, or roots with a radical vowel, and one should perhaps not deny categorically that some of these roots may de- rive, via processes thus far not well understood, from other root types in some early stage of Proto-Semitic, or in Afro-Asiatic. But, in my view, the developments in the descendant branches and individual languages can be explained much more economically and regularly, and reconstruc- tion proceeds much more smoothly, if we assume triconsonantal roots at least in the latest stage of common Semitic, just prior to the initial divi- sion into East and West Semitic 47 7. The Prefix-conjugation of Geminate Verbs Geminate verbs are subject to special phonotactics in nearly all of the Semitic languages, and some of these special features are no doubt to be reconstructed to Proto-Semitic. For example, in both Akkadian and He- brew, the only adjectives that have the pattern qatt% are those of stative geminate roots, such as Akkadian dann- and Hebrew *qall-. The fact that this unusual aspect of geminate roots appears in both Akkadian and Hebrew suggests that it is inherited from Proto-Semitic,49 probably the result of a Proto-Semitic vowel syncope rule (i.e., *danVn- > *dann-).50 The scholarly literature on this topic is very large. For the triconsonantal view, see, e.g., GVG 1.632; see also Voigt 1988. 48 Or, the only verbal adjectives with a monosyllabic base, apart from those from hollow roots. 49 See Meillet 1925, 27, for the principle that shared paradigmatic anomalies are generally to be reconstructed to the proto-language. 50 In other words, internal reconstruction suggests a deviation from a regular non-geminate pattern qatvl-. On the forms in question see Huehnergard 1987b, 230-231, n. 59. Cf. also the suffix-conjugation forms of geminates in other languages (Brockelmann, GVG 1.257 §96a), such as Arabic dalla, Aramaic cal(l), Ugaritic forms RjR.2 (normally, e.g., sb = /sabba/ 'he turned'), and perhaps Mehri cfoZZ (unless this paradigm is from Arabic); Gs'sz nadda, etc., com- prise a subset of the gabra type. 172 JOHN HUEHNERGARD Another noteworthy feature of geminate roots is the well-known clus- tering of the geminate radicals in prefix-conjugation forms, as in Arabic yadullu, and in the Hebrew plural forms yosobbu, yeqdllu, instead of **yisb3bu, **yiqlalu. In Ugaritic, too, in prefix-conjugation forms of the G stem of geminate roots, overwhelmingly only two root consonants are written, as in ygz = /yaguzzu/ 'he cuts'.51 Similar forms are probably also to be reconstructed for early Aramaic,52 but later changes on the analogy of verbs \-n have generally obscured this phenomenon. Internal reconstruction suggests that the paradigm of these verbs in Cen- tral Semitic reflects a sound change; given, for example, yaqtulu, the form *yasubbu presumably derives from earlier *yasbubu; in other words there is, vis-a-vis the sound verb, a regular metathesis of the sec- ond radical consonant and the theme vowel.53 The sound change in ques- tion may be described as follows: C1C2V1C2V > C1V1C2C2V, or, CiV > VCi / C_CiV; thus, *yasbubu>*yasubbu *yiqlalu > *yiqallu. The corresponding forms outside of Central Semitic do not exhibit this Hebrew forms such as sabboto (cf. haqimoto), Arabic maddata are also probably inherited from Proto-Semitic rather than directly from Proto-Central Semitic. The *-a- before the suffix is clearly cognate with that in Akkadian pars-a-ku, etc. 51 So also in imperative forms, such as sb = /sub(bv)/ 'turn'. See Tropper 2000, 672-679 §75.6. 52 Note perhaps the Biblical Aramaic G (pacal) imperative goddu 'cut down' (Dan 4:11,20) and the C - (hopcal) perfect hucdllu (Dan 5:15); Bauer/Leander 1927, §§48c,k, too hastily at- tribute these to Hebrew influence. The lack of vocalization in Old Aramaic inscriptions prevents us from knowing whether forms such as imperfective ycl still had the shape /yicul(l)/ or had shifted to the 1-n pattern, /yiccul/; for the forms see Degen 1969, 72-73. 53 Compare Brockelmann, GVG 1.258 §96c: "Formen wie iaruddu ... konnen aber auch erst aus den spateren Formen wie Hardudu entstanden sein; dann ist die unbequeme Aufeinander- folge der gleichen Silben durch Metathesis der 1. beseitigt worden." Brockelmann goes on to note Arabic Qur'anic variants of form VIII (Gt) verbs that exhibit a similar sound change (ac- companied by assimilation of the t infix), as in yahtatifu > yahattifu, taHadu > tacaddii. FEATURES OF CENTRAL SEMITIC 173 feature; they are quite regular, as in Akkadian idninu™ Ga'az yahsasu,55 MSA Mehri subjunctive yadlel, 2fs tddleli.56 Thus, the Cen- tral Semitic languages pattern together in this regard over against "non- CS" languages. That we have to do here with a sound change is suggested by forms with the same syllabic structure in the causative stem in Hebrew and Arabic.57 Compare: PNWS *haqtila ~ *yuhaqtilu, but *hasibba ~ *yuhasibbu5S Arabic Daqtala ~ yuqtilu, but Daqalla ~ yuqillu. In these forms, too, there is a metathesis of the second radical consonant and the following vowel. Again, Gs'sz exhibits regular forms such as 54 Rare Neo-Assyrian forms such as (Sa) ak-kar-ru-u-ni /akkarr-u-ni/ 'which I threw', for normal akrur-u-ni, undoubtedly reflect Aramaic influence; see GAG § 101 f; Kienast 2001,345. 55 But note the occasional G imperative hassu beside hasasu (Dillmann 1907, 105 §55; Brockelmann, GVG 1.635 §272Fa) and G imperfective yandddu alongside yanaddadu; in both of these cases we probably to have to do with the loss of a vowel or syllable rather than the con- sonant-vowel metathesis seen in Central Semitic. These forms are curious in that, synchronically at least, it is an accented syllable that is lost. 56 The imperfective of these verbs has the form yddlol, rather than the expected %*ydddldl (Johnstone 1987, xxiii); Harsusi forms are similar: del Iyedlol Iyedlel (Johnstone 1977, 24). In Jibbali too we find ydfrer; the subjunctive forms double the first radical, yoffer (cf. Aramaic *yiccul, etc.); note, however, the 1-guttural forms, imperfective ycazez, subjunctive yaczez (Johnstone 1981, xix). 57 Note also the suffix-conjugation of the N-stem form in Hebrew: *naqtala, but *nasabba. In Ugaritic, however, we seem to have N forms of the suffix-conjugation both with and without the change: nsb = /nasabba/ and nSdd /naSdada/ (see Tropper 2000, 677). The latter may be a relic form or a new formation based on the analogy of the sound root; see further below. (In the PNWS N prefix-conjugation, the original syllabic structure is different, in that the first and sec- ond radicals are not contiguous, and the sound rule does not apply, since a metathesis of R.2 and the vowel between R2 and R3 would result in a sequence of two vowels; rather, there is vowel syncope: *yissabibu > *yissabbu; so also in the Arabic N forms of both the suffix- and the prefix-conjugation: inqatala ~yanqatilu, but inqalla ~yanqallu.) 58 Ugaritic causative forms of geminate roots are unfortunately rare; Tropper 2000, 679 lists two possible examples, one of which exhibits the sound change, ySlh /yusalih(hu)/, and one of which does not, y&hmm IyuSahmimu/. As with the longer N stem form cited in the preceding note, yShmm may be either a relic or an analogical re-formation. 174 JOHN HUEHNERGARD 3ahsasa and yahsdsu; in Akkadian, too, we find regular forms such as suklulu, usaklilu.59 This consistent treatment of prefix-conjugation forms of geminate verbs in Ugaritic, Hebrew, and Arabic is striking. So too is the fact that the metathesis of a consonant and a vowel is, it would seem, a rare phe- nomenon in Semitic (unlike, e.g., English bird < brid, etc.).50 These fac- tors argue strongly that these forms are inherited from a common ances- tor, rather than the result of parallel development or of an areal phe- nomenon. The fact that such verbs are regular in Akkadian and Ethio- pic61 might further suggest that they reflect a shared innovation in Proto- Central Semitic. There are however, problems with such an interpreta- tion. There are exceptions in each language that give one pause, forms that are patterned on the sound verb.62 Examples are Hebrew Hiphil forms such as tarnin instead of **toren;63 in the Qal, instead of the usual yohon, we find once yeh£nan (Amos 5:15), a form that matches yi-ih-na-nu(-ni) in a Canaanizing Amarna letter.64 In early Arabic dia- In MSA Mehri, some forms, at least, have the geminate radicals together: 3fs hadallut, vs. 3ms hddlul, 2ms hadlek < hadlalk (with vlC > vC as elsewhere; hadlalk occurs in southern Mehri, according to Johnstone 1987, xxxix). 3ms hddlul may reflect earlier *hadlala> as in Ethiopic; but it is also possible that it reflects a phonological distribution as in Arabic dalla(t), dalalta, i.e., with geminated consonant only before vocalic suffix, and CVC otherwise. But is 3fs hdddllut the result of a sound change as in Central Semitic, or is the paradigm influenced by Arabic? 60 On the metathesis of CV > VC, see Blevins/Garrett 1998. 61 Although in Mehri and Harsusi the G yaqtul forms of these verbs are for the most part as in Gs'sz and Akkadian, the partial patterning of the C suffix-conjugation forms as in Central Se- mitic requires additional investigation, as do the Aramaic-like forms of most G yaqtuh in Jibbali (see above, nn. 56, 59). It would seem that a number of internal-MSA changes have occurred. 62 We are not concerned here with Hebrew forms with doubling of the first radical, such as yissob; the source of the doubling of the first radical in such forms is debated, but most agree it is secondary. See, for example, Bauer/Leander 1922, 433-434; Brockelmann, GVG 1.635-636; Kautzsch 1906; Muller 1986. 63 All causative forms of the root r-n-n are regular (GCK §67cc). 64 EA 137:81, from Byblos. FEATURES OF CENTRAL SEMITIC 175 lects there are also rare examples of yadlulu rather than yadullu.65 Ug- aritic, too, may attest a few examples patterned on sound verbs.66 And, finally, in OSA Sabaean we find prefix-conjugation forms in both the "geminate" and the sound verb pattern, and causative forms are some- times in the "geminate" pattern, but more often in the sound verb pat- tern.67 Are the "non-geminate" forms in these languages relics, that is, archaisms or holdovers that for some reason did not undergo the sound change? Or are they innovations, "non-paradigmatic" forms, as it were, created spontaneously and independently in each language on the anal- ogy of the sound verb?68 The latter is entirely possible. And that suggests that the sound conjugation of such verbs in Akkadian, Ethiopic, and MSA may similarly be the result of secondary analogy with the sound verb, independently in each language (group). In other words, it seems that we do not have any firm basis to conclude that the yasubbu-type conjugation and the sound rule that created it were not already features of Proto-Semitic, features then lost by analogy or levelling independently 65 E.g., nirbabu-hu 'we rear him'; see Rabin 1951, 90, 161; Hopkins 1984, 76. 66 Further, in both Hebrew and Ugaritic, verbs that are both geminate and l-n may be conju- gated like other \-n verbs, as in Hebrew yiddod, not *yonod. (But note Hiphil yniddtihu, a geminate form, alongside the l-n type Hophal yuddad.) For Ugaritic, see Tropper 2000, 675 §75.62c. If the assimilation of n is a relatively late phenomenon (PNWS), or at any rate later than the appearance of the yasubbu forms of geminate verbs (whether that is PS or PCS), we should expect yanuddu to have been the inherited form; thus, yiddod would reflect a repatteming on the analogy of other verbs l-n. 67 See Nebes/Stein 2004, §4.4.2.2, 4.4.3; Nebes 1994. Forms of geminate roots culled from Beestonetal. 1982: G prefix-conjugation: "geminate" pattern: ymrn; ygln; ysh ; "sound" pattern: yHmmw; ybrrn (unless these are D or L forms); Causative verbs: "geminate" pattern: yhbQ (vs. G bddt 'announce'); hrdy (dual); hrs1; hsr; "sound" pattern: hclln; hcmmn; mhcmm (mimation?); hczz\ hbrr(w), yhbrrw; hdll, yhdlln; hgdd; yhglln; hrbb; hs2nn. 68 Biforms such as bozdznu Ibazzonu, zomdmti I zammoti also suggest repatteming on the basis of the sound verb; the second form in each of these pairs almost certainly derives from a Proto-Semitic ancestor. 176 JOHN HUEHNERGARD in Akkadian, Ethiopic, and MSA,69 and sporadically, as we have just seen, in the Central Semitic languages as well. 8. The Suffix-conjugation of Verbs U-w/y The paradigms of the suffix-conjugation of middle-weak verbs in He- brew, Aramaic, and Arabic may also suggest a common subgrouping. Paradigmatic forms of the third and second persons masculine singular appear below. Unfortunately we know nothing about the vocalism of these verbs in the OSA dialects. In Ugaritic, there are still no examples of suffix-conjugation forms from roots with an aleph in first position, so there too the vocalism remains unknown. 70 Akkadian Gs'az Arabic Hebrew71 Aram. (Bib/Syr) kin~kinata qoma~qomka qdma~qumta qom~qdmto qam~qamt Slm-Simata sema-semka Sama-Simta som~sdmto sam-samt mit~mitata mota~motka mata~mitta met~mdtto mit~mitt tab~tabata taba~tibta tob_~H6bto (tOeb-tOebt) So also Brockelmann, GVG 1.634, 638. The analogies with the sound verb would perhaps be more obvious in these languages, in which the imperfective form of the geminate roots gener- ally retains the shape of a sound verb; in Akkadian, e.g., given idannin, analogies such as irappiSu : irpiSu :: idanninu : X = idninu would be obvious and would perhaps more easily replace *yadinnu (similarly in Gs'az with, e.g., ymaddaf : yzndaf :: yanaddad : X = yandad). 70 Verreet apud Tropper 2000, 642 §75.521c, notes forms such as lcs/2ms mtt and Stt, and plausibly suggests that these may indicate a vowel before the suffix, i.e., /sTtatu/ or /satatu/, as in Hebrew G riboto, N nqumoto, C haqimoto; see further the next note. 71 Note also Hebrew II—y forms such as bin, binoti beside bdnto; riboto beside rdbto (GKC §73a). Despite the fact that they are mostly late, it is unlikely that they are Aramaisms, in view of the presuffixal -6- (< *a). The more frequent appearance of the -6- in C and N forms (haqimoto, nsugoti), as well as the possibility that similar forms occurred in Ugaritic (see the preceding note), suggests the further possibility that such Hebrew forms may be relics of an archaic, parallel conjugation of these roots in the suffix-conjugation, similar to Akkadian klndta, etc. FEATURES OF CENTRAL SEMITIC 177 Mehri: mot~matk(3fs matut); {teyob-pydbk) Note as well the Amarna Canaanite forms nu-uh-ti, i.e., /nuhti/ 'I am at rest',72 which parallels Arabic qumtu, and G passive si-ir-ti, i.e., /girt!/ 'I was slandered',73 corresponding to a sound verb pattern qatilti and to Arabic qiltu.74 These forms indicate that in at least some Northwest Semitic dialects these verbs may have had an Arabic-like con- jugation. Further, application of Hetzron's principle of archaic heteroge- neity leads us to conclude that conjugations like those of Arabic, with their different vowel qualities and quantities within the paradigm, are the more original.75 Finally, if we take such heterogeneous paradigms, i.e., *qama~qumt§, and *sdma~s(mta, to be earlier, the attested Hebrew and Aramaic paradigms of these verbs can be derived by a series of sim- ple levellings: in both Hebrew and Aramaic, the allomorphism of the vowel quality was first levelled, in favor of that of the third person forms: thus, e.g., *qama~qumta became *q&ma~qamt5,. Hebrew and Aramaic then went their separate ways, Hebrew levelling the short- vowelled allomorph, *qdma~qdmta, and Aramaic levelling the long- vowelled allomorph, *qdma~qdmta. 'To die' was probably i-class in the suffix-conjugation originally, i.e., *mlta~mttta,76 levelled in He- brew to *mtta~mttta, in Aramaic to *mUa~mitta, while in Arabic the 72 EA 147:56 (from Tyre). 73 EA 252:14 (from Shechem). 74 That Si-ir-ti is passive is confirmed by the prefix-conjugation (1st comm. sg.) form /°usaru/ (written u-Sa-a-ru, u-Sa-a-ru, EA 286:6,21,24 [from Jerusalem]), corresponding to the sound pattern yuqtalu and to Arabic yuqdlu. 75 Also a factor in favor of taking the Arabic forms as archaic is the phonology. The pairs *qdma~qumta and *s&ma~simta can be derived from still-earlier, "regular" *qdwama~qawdmta, *sdyama~saydmt3. We may posit the following Proto-Semitic sound changes: diva, dya >al _CV, but awd > u, ayd >i I _CC; the PS status of these changes is also indicated by Akkadian forms such as tabum 'good' < Hdyabum and tdbbal 'you bring' < *tawdbbal. For the Hebrew forms see also the insightful article of Holmstedt 2000, esp. 149-152. 76 From still-earlier *mdwita~mawttt5. 178 JOHN HUEHNERGARD third person form was replaced by mata on the analogy of verbs such as sama-simta. If this reconstruction of the development of the paradigms in Hebrew and Aramaic is correct, then it follows that we must posit a common an- cestor for the paradigm for Hebrew, Aramaic, and Arabic. The question then becomes: Is this common ancestral paradigm of the middle-weak suffix-conjugation an innovation or a retention? Since the Akkadian and Ethiopic paradigms do not correspond with one another, it is difficult to determine how to reconstruct them back to Proto-Semitic. It may be sug- gested that the Akkadian forms, kin-, tab-, etc., may reflect the same phonological processes that yielded the Central Semitic forms;77 if so, then the Central Semitic paradigms may be descended from something quite similar in Proto-Semitic, and should thus be considered a shared retention rather than a shared innovation.78 The Mehri forms have been left out of consideration here, because their earlier shape is ambiguous. They could be the result of a series of developments as in Hebrew,79 or a series of analogies to the sound verb. Additional research is needed. 77 Viz., *kawin- > *kln-, Hayab- > *tab-1 tub-, with the long-vowel allomorph general- ized in each case. 78 The Ethiopic forms would then be the result of a change or set of changes in Proto- Ethiopic, viz., a generalization of theme-vowel i in a fully-triconsonantal base: *qawima, *sayima; with the general loss of the second vowel in such forms (*kabida > kabda), these forms became first *qawma, *sayma, and finally qoma and sema with the reduction of the diphthong. (For the intermediate forms *qawma, *sayma, note verbs such as sawca and haywa.) The rest of the paradigm was presumably then levelled on the analogy of the sound verb: nabara : nabarka :: qoma : X = qomka. 79 I.e., the Mehri forms listed could be from *mata~matat(3fs)~matka; alternatively, mot could also be from earlier *mata or even *mdta. The 3fs could be the result of analogy with the sound verb: rakdz : mot:: rakazot: X = matot. The vowel of 2ms matk is difficult in view of sound nkazk. FEATURES OF CENTRAL SEMITIC 179 9. The Change of Verbs III-u; to III-/0 Certain verbs with third radical w in Ethiopic, and originally in Ak- kadian, exhibit either partial or entire paradigms with y in the various Central Semitic languages.81 As already suggested by Brockelmann, this was probably at least partly the result of a sound change, namely, w>y I ■ y 82 jjius wg £n(j jn Aj.akjc ver|3s 0f the type radiya; the finite conju- gation of this verb is entirely lll-y, but the masdar, ridwan- 'consent', and comparative evidence show that the root was originally III-u;; thus, first radiwa (*rasiwa) became radiya, and then the rest of the finite conjugation was levelled to lll-y. In the verb Jata~yaDti 'to come', which Ethiopic Datawa shows to have been III-u; originally, presuma- bly of the a~i class, the sound change would have operated in the prefix- conjugation, with original *yaDtiwu becoming *yaJtiyu, the y then be- ing levelled into the suffix-conjugation as well. This verb has shifted en- tirely to a III-y root in Arabic, including the verbal noun Dityan-, al- though there is also a derived noun Htawa 'tribute'. The intermediate stage of this process can be seen in an interesting pair of Ugaritic forms, suffix-conjugation atwt = /^atawat/ and prefix-conjugation tity = /ta3tiyu/;83 here too an original prefix-conjugation form *taDtiwu has become /ta3tiyu/ by the sound change, but the suffix-conjugation form has not yet been levelled through. The continued existence of forms such as the Old Aramaic (Sefire) D imperative (m. sg.) rsw = /rassiw/, the Moabite D yaqtul 3ms ynw = /yiranniw/, and the Hebrew adjective -onow means that the sound change was occasionally blocked by paradigm pressure. In the OSA languages we also find a number of forms of roots III—^ that exhibit a As noted above, presence of final consonantal radicals w and y is assumed in the follow- ing; thus we reject the conclusions of Diem's classic 1977 study of Ill-weak verbs. 81 In most modern Ethiopian languages (though not Tigrinya) these root types have also fallen together, via a similar, but unrelated set of developments. 82 Brockelmann, GVG: 1.618-619; see also Hopkins 1984, 84 on the frequency of the change of verbs III-h; to III—y within Arabic. 83 See Olmo Lete/Sanmartm 1996-2000, 59. 180 JOHN HUEHNERGARD final y instead; not infrequently, these occur alongside forms with w, also attesting to the tension between a sound change and paradigmatic pressure. Since there is no evidence for the sound rule outside of the Central Semitic languages,85 it must be either a shared innovation from Proto- Central Semitic or, perhaps less likely, an areal phenomenon. In Hebrew and in Aramaic, of course, the distinction between roots originally UI-w and roots originally III-j has for the most part been completely obliterated, usually to the forms of the latter, as in Hebrew hoytto (Ga'az haywa); yortti (Ga'az warawa).86 This was the result of levelling analogies following other sound changes,87 but perhaps also partly the result of levelling following the sound change posited here. 10. Barth-Ginsberg Law In 1894, J. Barth published his famous observation that, in the prefix- conjugation forms of certain types of Hebrew G (Qal) verbs (I-guttural, geminate, II-weak), the vowel of the prefix varies with the theme-vowel, viz., with theme-vowel u and i the prefix has a, while with theme-vowel a the prefix has i, i.e., we find yaqtul, yaqtil, but yiqtal.88 Barth also suggested that such a distinction also underlies the West Syriac imper- fects of verbs 1-, such as nekol 'he will eat' vs. nlmar 'he will say'; vestiges of prefixes in -a- also appear in verbs I-guttural in Biblical Aramaic and in the unique ya-bed of Targum Onqelos. H. L. Ginsberg established that the same distribution of prefix vowels obtains for all G 84 E.g., yHvon, yDtyn 'he will come', yhrdwn, yhrdyn 'he will satisfy'. See Nebes/Stein 2004, §§3.4.1,4.2.2(2). 85 The trend in the history of Akkadian for all Ill-weak roots to be levelled to the III—i type is probably not due to the effect of a sound change, but rather to a general reduction of types. 86 Note also, e.g., rosuy; further, with archaizing preservation of the third radical: yeDEt5yu, ye^toyun. 87 E.g., in Proto-Hebrew, final *-awu, *-ayu, *-iyu > -e and final *-awa, *-aya, *-iya > *-a > -o. 88 Barth 1894,4-6. FEATURES OF CENTRAL SEMITIC 181 verbs in Ugaritic. The phenomenon has also been noted vestigially in classical Arabic and in certain Arabic dialects.90 It is likewise exhibited by one or two Amarna Canaanite forms.91 Finally, it has been suggested that the same distribution of prefix-vowels may account for the loss of y after the preformative I in certain OSA prefix-conjugation forms.92 It will be noted that all of these are Central Semitic languages; there is no evidence of this phenomenon in Ethiopian Semitic, in the MSA lan- guages, or in Akkadian.93 Barth considered the distribution of prefix vowels to reflect the origi- nal Semitic situation.94 It is more likely, however, that Hetzron was cor- rect in considering the heterogenous paradigm of the prefixes in Ak- kadian, where we find ta-prus with -a- but ni-prus with -i-,95 to be the more primitive.96 We may, therefore, suggest that the Barth-Ginsberg "law" is a common Central Semitic innovation.97 89 Ginsberg 1932, 382-383; 1938, 318. 90 Bloch 1967; Schub 1973. 91 Rainey 1978; 1996, 61-75. Some of the examples cited by Rainey are graphically ambigu- ous, however; see Huehnergard 1987a, 720. 92 Hayes 1994. 93 A proposal by Testen 1992 to account for certain forms of Akkadian verbs l-w by invok- ing the Barth-Ginsberg phenomenon is not convincing. As will be noted below, the heterogene- ous nature of the Akkadian G prefix vowels (viz., ta-prus vs. ni-prus) is probably ancient, as suggested by Hetzron. An alternative view of the development of the l-w forms in Akkadian is presented in Huehnergard 1987d, 191-193. 94 Barth 1894, 6. 95 The third-person forms with initial i-, such as iprus, are ambiguous, since initial i- in Ak- kadian may derive from either *yi- or *ya- (for the latter, note, e.g., idum 'arm' < *yadum, iSarum 'straight' < *yaSarum). 96 Hetzron 1973-74, 35-40; 1975; 1976, 94-95. For an alternative approach, see Izre'el 1991. 97 So also Hetzron 1973-74, 40. I wish to thank my student R. Hasselbach for sharing with me a draft of her paper on the prefix vowels which prompted me to rethink my views on this issue; see now Hasselbach 2004. 182 JOHN HUEHNERGARD 11. G passive *yuqtal A G passive prefix-conjugation pattern yuqtal is attested in Amarna Canaanite (e.g., tumhasu 'it (f.) is smitten'),98 in Biblical Hebrew (e.g., yuqqah 'it will be taken'), and in Arabic. Passive yqtl also occurred in Ugaritic," Old Aramaic,100 and OSA;101 the lack of vocalization makes it impossible to ascertain the precise shape of the form in those languages, but it seems quite likely that *yuqtal may be reconstructed at least to Proto-Central Semitic. Ethiopian Semitic exhibits no finite internal pas- sive forms. In Akkadian, too, the only internal passives are the verbal adjectives, *qatil-, and the predicative construction based on them; no prefix-conjugation forms of the *yuqtal type occur. These factors might suggest that passive *yuqtal was a Central Semitic innovation.102 How- ever, similar forms are also attested in the MSA languages (as "subjunc- tives"): Mehri yarkoz 'may it be straightened'; Jibbali larfos 'may it be trampled'; Soqotri lerhos 'may it be washed'.103 These MSA forms are quite probably also reflexes of earlier *(lv-)yuqtal, which should thus be posited for Proto-MSA. It is possible that the MSA *yuqtal has been borrowed from Arabic. If not, however, then it would seem that *yuqtal is not an innovation of Proto-Central Semitic, but rather either an innova- tion of Proto-West Semitic which was lost in Proto-Ethiopic, or an inno- vation of a common intermediate ancestor of Central Semitic and Proto- MSA, after Proto-Ethiopic had branched off. 12. The Form of the'tens' In Akkadian and in classical Ethiopic, the words for the 'tens' are 98 See Rainey 1996,2.75-80. 99 See Tropper 2000, 510-514. 100 Degen 1969, 66, 69. 101 SeeBeeston 1984, 14 §5:3; Nebes/Stein, §4.4.1. 102 Cf. Blake 1901, esp. 47. 103 For these forms see Johnstone 1975, 19; 1981, xvii-xviii; 1987, xxi-xxii; Simeone-Senelle 1997,407. FEATURES OF CENTRAL SEMITIC 183 formed by adding a long -a to the base of the corresponding unit: for ex- ample, '30' in Akkadian is salasa and in Gsc9z it is salasa.104 The 'tens' in the MSA languages seem to be similar in formation; according to Simeone-Senelle, "The tens, when not borrowed from Arabic, are made by suffixation of -ah, -oh, -0h" as in the word for '20' in one Mehri dialect, -dsardh, and in one Soqotri dialect, -dsr0/i.105 In Hebrew, Aramaic, and Arabic, however, the 'tens' are formed as external mascu- line plurals of the corresponding units, as in Hebrew slosim, Aramaic t(d)latin, Arabic Oaladuna. In Ugaritic, the unvocalized nature of the script makes it difficult, perhaps impossible, to know the shape of the 'tens' with certainty, but the fact that they exhibit a final -m indicates that the ending was not a simple vowel as in Akkadian and Gs'sz. In- deed, Tropper in his recent grammar of Ugaritic has also noted that the formation of the 'tens' as plurals is a feature of Central Semitic, and ac- cordingly suggests that the same was true in Ugaritic.106 This morpho- logical change is peculiar enough that it is probably to be considered a shared innovation inherited from a common ancestor.107 In the Old South Arabian languages, the 'tens' exhibit a final -y, as in 9l9y for '30'.108 A final -y marks the bound form of both the dual and The origin of these forms, whether a dual ending or an absolute nominal ending, and whether the Akkadian and the Gs'sz forms are cognate, was debated by W. von Soden (1961; 1983) and M. Powell (1981). 105 Simeone-Senelle 1997, 396. 106 Tropper 2000, 354 §62.3. 107 On the other hand, the higher numbers would probably have been used especially by mer- chants and traders, and so might well have traveled in wave fashion. Another possibility, sug- gested orally by J. Blau, is that originally '20' was the dual of '10' and that '30' through '90' were plurals of the corresponding units; the attested paradigms would then simply be the result of levelling to either the dual or the plural throughout (so also Reckendorf 1894). This is an attrac- tive suggestion. It is, however, not clear that the forms of '30', etc., would have been plurals originally. It is also not certain that the ending -a in Akkadian and Ga'sz is an old dual marker; the latter claim has been sharply criticized, e.g., by von Soden (1983). 108 In Minaean and Hadramitic, final -hy is also attested, as in Minaean Drbchy '40' and Had- ramitic tslchy '90' (Kogan/Korotayev 1997, 232). 184 JOHN HUEHNERGARD the external masculine plural, so that if the 'tens' are construed as bound forms before the nouns they quantify, it is not possible to determine whether the forms are plurals as elsewhere in Central Semitic, or instead are cognate with the final -a of Akkadian and Ga'az.109 Beeston in his Sabaic Grammar notes a Qatabanian example that "tends to support the hypothesis that the decad numerals are not syntactically construct";110 if that is true in that one instance, and true in general in Qatabanian and in the other languages, it would suggest that the OSA 'tens' resembled those in Akkadian and Ga'az rather than the plural-like forms of Hebrew, Aramaic, and Arabic, and, thus, that the OSA languages did not partici- pate in this innovation that is shared by the other Central Semitic lan- guages. That, in turn, would indicate that Northwest Semitic (Ugaritic, Canaanite, and Aramaic) plus Arabic constitute a subgrouping within the larger Central Semitic group; but the evidence here is admittedly thin. 13. Definite Article It seems unlikely that a definite article is to be reconstructed for Proto-Semitic; it is lacking in Akkadian and in classical Ethiopic, and has a wide diversity of forms and syntax in the languages in which it is attested, factors that suggest later innovation in those languages. A defi- nite article appears in the Canaanite and Aramaic dialects from their ear- liest attestation,111 in pre-classical and classical Arabic and the later Ara- bic dialects, in the Old South Arabian languages, in most of the modern Ethiopian languages, and in at least some of the MSA languages. In Ethiopian Semitic, the diversity of the morphology of the article in the As noted above, n. 43, there are a number of instances in some of the OSA dialects in which a final -y appears on forms in which, from a comparative perspective at least, *-a is ex- pected. 110 Beeston 1984: 65, §Q 18:5. 111 Except for Amarna Canaanite. As noted by Lambdin (1971, 318-319), the article in the earliest Aramaic inscriptions is not common, and with few exceptions is restricted to a small number of syntactic environments - in other words, its occurrence is essentially syntactically conditioned rather than semantically. FEATURES OF CENTRAL SEMITIC 185 modern languages - independent demonstratives in Tigrinya, prefixed la- in Tigre, suffixed third person pronouns in Amharic and Harari, etc. - indicates a late, i.e., post-Proto-Ethiopic origin, unrelated to the devel- opment of the article in other branches of Semitic. The article in the MSA languages is still not well understood, and its origin remains un- clear, but it differs in its syntax from that of the Central Semitic lan- guages, for example, in appearing on nouns with suffixes, as in Mehri a- bayt-i 'my house' (Johnstone 1970). The article in the Canaanite languages and in Arabic almost certainly has a common morphological origin. Whether the post-positive Aramaic article has the same or a similar morphological origin, as has been ar- gued by a number of scholars, remains uncertain, as does the origin of the article in the OSA languages, a suffixed -n.U2 What is striking, how- ever, is the nearly identical syntax of the article in Canaanite, Aramaic, Arabic, and, probably, the OSA languages.113 Consider the following list of features: the article may appear only on the final member of construct chain; it may not appear on nouns with suffixes, or on proper nouns; ad- jectives must agree in definiteness when attributive; predicative adjec- tives are indicated syntactically, by the simple lack of an article in con- junction with a definite noun, as in Hebrew ham-melek fob 'the king is good', as opposed to ham-melek hat-tob 'the good king'.114 When we consider that most of these features are not inevitable, the fact that all of them characterize Canaanite, Aramaic, and Arabic is very suggestive of a set of developments in a common ancestor. On the other hand, it must be 112 See the recent studies of the definite article by Testen 1998 (ch. 4), Voigt 1998, Zaborski 2000, and Tropper 2001. 113 That is, the syntax of the article in the OSA languages, or at least in Sabaean, appears for the most part to be the same as that of the other Central Semitic languages. But more research is needed here; the grammars are not terribly clear about the syntax of the article in OSA. 114 Another syntactic commonality is the construction used for 'an X of the Y', exemplified by Hebrew ben lam-melek and Arabic ibnun li-l-maliki 'a son of the king' (see GVG 1.237- 238); it is also attested in Aramaic, e.g., Official Aram. Drdykl Iswn byrf 'a builder of Syene Fortress'; see Muraoka/Porten 1998, 215-218. The Gs'sz waldu la-nagus construction is probably unrelated. 186 JOHN HUEHNERGARD noted that not all of the languages that we would otherwise include in the Central Semitic group participated in this development: neither Ugaritic nor the Deir cAlla dialect, for example, exhibits a definite article. The absence of the article in those languages, as well as its relative rarity in some early Hebrew poetry and in the earliest Aramaic inscriptions, and its diversity of forms not only across languages but also, for example, in early inscriptional dialects of Arabic - all of these factors would nor- mally indicate that we are dealing here not with a shared innovation in- herited from a common ancestor, but rather with a wave or areal phe- nomenon. It is something of a challenge to explain the wavelike aspect of the morphology of the article in Central Semitic alongside the shared- innovation aspect of its syntax in those languages. 14. Hebrew halloz(e) ~ Arabic alladl Biblical Hebrew exhibits a relatively rare demonstrative adjective of the form halloz, which is used to modify both masculine and feminine nouns. Even less often encountered is the specifically masculine form halloze, and only once do we find the feminine form hallezo. These forms are almost certainly cognate with the Arabic form alladl, a form that is no longer demonstrative, but has become the normal relative pro- noun.115 These forms are composed of the definite article,116 the Com- mon Semitic asseverative particle *la-, and the Common Semitic de- monstrative/relative pronoun *du. The combination is unusual enough that we may consider these forms to have originated in a common ances- tor. 15. ma-zzoOt'Dsito1? The common Biblical Hebrew expressions mi ze, ma-zze, ma- zzo(J)t, etc., have analogues in Arabic man da, ma da, as in wa-man da yafutu l-mawta 'wer kann dem Tode entgehn?' (GVG 2.196), ma- 115 Pace Bravmann 1977, 185-191. 116 Pace Rabin 1951, 155, who compared Amharic allazih. FEATURES OF CENTRAL SEMITIC 187 da tasna-u 'what are you making?' (Fischer 2002, 150). The construc- tion is also found in Syriac, where the forms man and manam proba- bly come from the fusion of the interrogative and the near demonstrative pronouns, that is, from ma + den Idna.n% These expressions are curi- ous. The demonstrative does not seem to be functioning as a copula, as, for example huD does in clauses such as -esow huD D£dpm 'Esau is Edom' (Gen 36:1); in the latter, the pronoun connects nouns or noun phrases, whereas with mi ze, etc., a verb may follow, as in mi-zs boD me-D£dom 'who has come from Edom?' (Isa 63:1), or ma-zze to(J)mru 'how can you say?' (Judg 18:24), or the common ma-zzoC)t -osito 'what have you done?' (e.g., Gen 3:13), or in the Arabic examples just cited. It is as though what follows is an asyndetic relative clause: ma- zzo(3)t -osito 'what (is) this you have done?'.119 Indeed, there are exam- ples in Arabic in which the relative pronoun also appears, as in man da lladl Jamara 'qui est celui qui a commande?', ma da lladi taqulu 'qu'est-ce que tu dis?' (Fleisch 1990, 2.76 §116f); there is a similar ex- ample in an Official Aramaic text,120 but there do not seem to be any Hebrew examples.121 There is also the fact that the Arabic da is other- wise rare as a demonstrative, at least, in the standard classical lan- 122 guage. It seems that this construction, in which a demonstrative is inserted Also °ayna, Dayda, Daylen; cf. Hebrew De-ze. 118 So Noldeke 1904,47; Brockelmann, GVG 1.326. 119 So also Zewi 2000, 52. Biblical Hebrew attests a number of asyndetic relative clauses; see Waltke/O'Connor 1990, 338. In standard classical Arabic, only indefinite antecedents are nor- mally followed by asyndetic relative clauses; see, however, Hopkins 1984, 240 §288, for exam- ples, in early Arabic papyri, of asyndetic relative clauses with definite antecedents. 120 mhy dh zy spr Ih hwSrtn ly 'What is this, that you have not sent me a letter?'; see Fol- mer 1995,578. 121 Unlike, e.g., in hP-lo^ ze ^Ser yiSte Dadoni bo 'Is not this that from which my lord drinks?' (Gen 44:5). 122 Fischer 2002, 145 §274 n. 2. Note the comment of Beeston 1984, 45 §28:4: "the origin of the Ar morpheme da in these cases [scil. man da and ma da] is equally obscure." 188 JOHN HUEHNERGARD immediately after an interrogative pronoun, is restricted to these Central Semitic languages. The construction may be compared with examples in Akkadian and Ga'az in which an interrogative is followed by a relative pronoun, as in the following examples: Akkadian: Old Babylonian mannum sa uqniam ana slmim inaddin-u? 'Who (is it) that will sell lapis?' Birot 1993, 267, text 161:14-15; Old Assyrian mlssu sa ammakam PN etawu? 'Why (is it) that PN has complained there?' Larsen 1988, 95, text 73:17-18. Ga'az: mdnta-ni za-wal<a>dku? 'What (is it) that I have borne?' Chaine 1909, 6:2-3 (Birth of Mary); wa-mannu za-ydkdl kama yahalli hdllinnahu, wa-mannu za-ydkdl nassarota la-kwdllu mdgbara samay? 'Who (is it) that can think his thoughts, and who (is it) that can observe all of the activity of heaven?' Knibb 1978, 352:23-23 (Enoch 93:11). In Akkadian this construction is infrequent, except for Old Assryian mlssu sa {GAG §47b). In Ga'sz, however, it is quite common.123 In the Akkadian and Ga'az examples, the relative pronoun essentially produces a cleft sentence,124 a common construction with interrogative pronouns in many languages that adds a certain degree of emphasis or vividness:125 instead of 'who did X?' we have 'who is it that did X?'. This construc- tion, with a relative pronoun, seems to be rare in the Central Semitic lan- guages;126 it may be suggested that the Central Semitic construction with a demonstrative rather than a relative pronoun is a modification of the 123 Dillmann 1907, 517 §198e. 124 See Lambdin 1978, 128. 125 For the term "vividness" see Waltke/O'Connor 1990, 323. 126 A Hebrew example is mi MSsr loD colo ... 'who (is it) that did not come up...?' (Judg. 21:5; otherwise, mi Da$er means 'whoever'). In Official (Egyptian) Aramaic note the following example, with an anaphoric pronoun as copula: mn hw zy yqwm qdmwhy 'Who is the one who would stand before him?' (Muraoka/Porten 1998, 295 with n. 1155); see also the example cited in n. 120 above. FEATURES OF CENTRAL SEMITIC 189 latter, and, further, that the change is specific and unusual enough to be considered a shared innovation inherited from a common ancestor. It is tempting to suggest that the demonstratives in the Hebrew and Arabic constructions, especially in the latter, are actually frozen or grammaticalized forms of the old Semitic relative pronoun, *dv,ni which was otherwise modified or lost for the most part in those lan- guages, being replaced by allaSl in Arabic and 3(Xser (and se+) in He- brew. In Old South Arabian, Sabaean inscriptions attest a few examples of an interrogative pronoun (mn/mhn) with a following 5- or 1-, which correspond semantically to Arabic relatival man and ma. It has been suggested that the second element is a relative pronoun.128 If that is the case, then these examples are like those found in Akkadian and Ethiopic, and this would then constitute another feature that Sabaean, at least, does not share with the other Central Semitic languages, and further indication of an internal subgrouping.129 It should also be noted here that the form of the impersonal interroga- tive, *mah-, also groups the Central Semitic languages together, as op- posed to Akkadian and Ethiopic *min(-t)-,m as has been pointed out by others. It seems improbable that two forms are to be reconstructed to Proto-Semitic.131 16. Lexicon Mention of the form of the interrogative pronoun leads to another is- See Muraoka 1985, 134-137, for discussion and bibliography. 128 Ryckmans apud Beeston 1984, 45-46 §28:4 with n. 80; Beeston himself was hesitant to accept Ryckman's explanation, however. 129 Note, however, that interrogatives are extremely rare in OSA thus far, and so there is little opportunity for phrases of the ma-zzo(D)t type to occur. 130 Note, however, that the early Akkadian particle ma sometimes has the force of an in- terjection expressing doubt or disbelief, rendered 'what?!' by the dictionaries (AHw 570b; CAD M/l 1). 131 Pace Blau 1978,34. 190 JOHN HUEHNERGARD sue, namely, the lexicon in general. Lexical items pose special difficul- ties, for the most part because it is not clear what it means when one sub- set of languages exhibits a lexeme that is not found in another subset. It is well known, for example, that the word *-abd- 'servant, slave' is re- stricted to the Central Semitic languages.132 There are scores of such words and roots, and even some grammatical elements, such as the prepositions *Dilay- and *-imm-, and the conjunction *pa- 'and (then)'.133 But it is usually possible - and probably often correct - to sug- gest that such lexemes are inherited from the proto-language and were simply lost in the languages in which they fail to appear. (It must be ad- mitted, however, that such reasoning also leads inevitably to the unlikely conclusion that the lexicon of the proto-language must have been larger than that of any of its descendants.) It is also possible to find roots and words in subsets of languages that would indicate other subgroupings. Furthermore, it is often difficult to rule out borrowing, especially among closely related languages, where loans can be notoriously hard to detect.134 What would perhaps be valuable for possible genetic subgrouping would be shared semantic shifts that are in some respect unusual. For instance, if there is a root in most of the languages, and it has meaning X in Akkadian and in Ethiopian Semitic, but has meaning Y, which is a likely but unusual semantic shift from X, in the Central Semitic lan- guages, that might indicate that the semantic development occurred in a common ancestor. But it must be admitted that there would be a signifi- cant chance of borrowing or areal spread in such instances, as well.135 132 The alleged Eblaite cognate **cebdu in personal names is almost certainly to be inter- preted otherwise; see Krebernik 1988, 37; Conti 1990, 29. 133 The related form *Dap, however, also occurs in Eblaite; see Fronzaroli 1981. For Central Semitic *pa-, see Althann 1997, ch. 5. 134 See again Hetzron 1976. 135 See, e.g., Schall 1982, 144, on Arabic roots such as kataba that have an indigenous meaning ('to sew together') and one that has been borrowed from Aramaic ('to write'). FEATURES OF CENTRAL SEMITIC 191 We might also look for shared morphological changes in the lexicon, such as the extended form of the word for 'god', *Jil-dh-, which is found only in the Central Semitic languages,136 and is probably a back- formation from the plural *Dil-ah-uma,ni itself a double plural of the type seen, e.g., in Syriac Dabahata and Dabahe 'fathers' and smahata and smahe 'names'. Again such shared forms could reflect either inno- vation in a common ancestor or areal spread. III. SUMMARY AND CONCLUSION Of the sixteen features we have reviewed, only a few can be said with some confidence to reflect shared innovations in a common Central Se- mitic ancestor; besides Hetzron'syaqtulu, these are: (1) the tense-mood-aspect system in general (10) the form of the 'tens' (11) the Barm-Ginsberg Law (14) Hebrew halhz(e) ~ Arabic alladi (15) ma-zzoC)t, etc. (and the form *mah-) Some of the other features may also examples of shared innovation, but we have seen that they may instead be retentions from common Se- mitic or instances of areal or wave spreading: (2) pharyngealization of "emphatic" consonants (perhaps areal) (3) *nihnu > *nahnu (perhaps areal) (4) feminine singular *-at > -a (parallel development) One might also consider unusual shared phonological developments such as metatheses, as in, perhaps, Common Semitic *sapc- 'foot, step' (Akkadian Sepum, Mehri saf, etc.) > *pasc- in Hebrew and Aramaic (see Kaye 1991a, 846-847); these too, however, could be areal phenomena. 136 For OSA, note especially Qatabanian sing. H vs. pi. 'IhnI'JlhwlDlhy; see Ricks 1989, 10. For Sabaean see Beeston et al. 1982, 5. 137 See GVG 1.334. 192 JOHN HUEHNERGARD (5) suffix-conjugation first-person ending *-ku —> *-tu (perhaps in- dependent development) (6) feminine plural *-na (probably common Semitic) (7) the prefix-conjugation of geminate verbs (perhaps common Se- mitic) (8) the suffix-conjugation of verbs U-w/y (probably common Se- mitic) (9) the change of verbs III-m; to III—^ (perhaps areal) (12) G passive *yuqtal (common West Semitic [then lost in Ethio- pic] or shared Central Semitic/Proto-MSA; if MSA forms bor- rowed from Arabic, then a Central Semitic innovation) (13) the definite article (areal?) (16) lexicon (common Semitic; areal/borrowing) We have also seen that, where there is evidence, the OSA languages also exhibit most of these developments. But there is some slight evi- dence that they did not participate in at least two of the shared innova- tions exhibited by the other Central Semitic languages, namely, the change in the form of the 'tens', and the innovative construction exem- plified by ma-zzoOt. If this evidence is valid, it suggests that Northwest Semitic and Arabic comprise a subbranch within Central Semitic, as in- dicated below in Figure 4. The two branches could perhaps be labelled North and South Central Semitic. Figure 4: The Subgrouping of Central Semitic Central Semitic North Central Semitic South Central Semitic (= OSA = Sayhadic) Northwest Semitic Arabic Sabaean Minaean Qatabanian FEATURES OF CENTRAL SEMITIC 193 REFERENCES AND ABBREVIATIONS AHw W. von Soden, Akkadisches Handwdrterbuch. 3 vol. Wiesbaden, 1965-81. Althann, R. 1997 Studies in Northwest Semitic. Biblica et Orientalia 45. Rome. Appleyard, D. L. 2002 New Finds in the 20th Century: The South Semitic Lan- guages, in: S. Izre'el (ed.), Semitic Linguistics: The State of the Art at the Turn of the Twenty-First Centwy (IOS 20; Winona Lake, Ind.) 401-430. Barth, J. 1894 Zur vergleichenden semitischen Grammatik. ZDMG 48, 1-21. Bauer, H. and P. Leander 1922 Historische Grammatik der hebrdischen Sprache des Al- ten Testaments. Halle. 1927 Grammatik des Biblisch-Aramaischen. Tubingen. Beeston, A. F. L. 1984 Sabaic Grammar. JSS Monographs 6. Manchester. Beeston, A. F. L., M. A, Ghul, W. W. Miiller, J. Ryckmans 1982 Sabaic Dictionary / Dictionnaire sabeen (English— French-Arabic). Louvain-la-Neuve/Beirut. Birot, M. 1993 Correspondance des Gouverneurs de Qattundn. Archives Royales de Mari 27. Paris. Blake, F. R. 1901 The Internal Passive in Semitic. JAOS 22, 45-54. Blau, J. 1968 Some Difficulties in the Reconstruction of »Proto- Hebrew« and »Proto-Canaanite«, in: M. Black - G. Fohrer (ed.), In Memoriam Paul Kahle (Berlin)... 29-43 = Blau (1998), 266-282. 1978 Hebrew and North West Semitic: Reflections on the Clas- sification of the Semitic Languages. HAR 2, 21-44 = Blau (1998), 308-332. 194 JOHN HUEHNERGARD 1980 The Parallel Development of the Feminine Ending -at in Semitic Languages. HUCA 51, 17-28 = Blau (1998), 126-137. 1998 Topics in Hebrew and Semitic Linguistics. Jerusalem. Blevins, J. and A. Garrett 1998 The Origins of Consonant-Vowel Metathesis. Language 74, 508-556. Bloch, A. A. 1967 The Vowels of the Imperfect Preformatives in the Old Dialects of Arabic. ZDMG 117, 22-29. Bravmann, M. M. 1977 Studies in Semitic Philology. Studies in Semitic Lan- guages and Linguistics 6. Leiden. CAD I. J. Gelb et al., The Assyrian Dictionary of the Oriental Institute of the University of Chicago. Chicago, 1956-. Cagni, L. (ed.) 1984 77 bilinguismo a Ebla. Atti del convegno internazionale (Napoli, 19-22 aprile 1982). Seminario di Studi Asiatici, Series Minor 22. Naples. Chaine, M. 1909 Christian, V. 1919-20 Conti, G. 1990 Degen, R. 1969 Diem, W. 1973 Apocrypha de B. Maria Virgine, Corpus Scriptorum Christianorum Orientalium, Scriptores Aethiopici, Series Prima, Tomus 7. Rome/Paris/Leipzig. Akkader und Siidaraber als altere Semitenschichte. An- thropos 14-15,729-739. II sillabario della quarta fonte della lista lessicale bilin- gue eblaita. Miscellanea Eblaitica 3 = Quaderni di Semitistica 17. Firenze. Altaramdische Grammatik der Inschriften des 10.-8. Jh. v. Chr. Abhandlungen fur die Kunde des Morgenlandes 38/3. Wiesbaden. Skizzen jemenitischer Dialekte. Beiruter Texte und Studien 13. Beirut. FEATURES OF CENTRAL SEMITIC 195 1977 1980 Dillmann, A. 1907 Dombrowski, 1988 EA Faber, A. 1980 1985 1990 1991 1997 Fischer, W. 2002 Fleisch, H. 1990 Die Verba und Nomina tertiae infirmae im Semitischen. Ein Beitrag zur Rekonstruktion des Ursemitischen und zur Entwicklung der Einzelsprachen. ZDMG 127, 15-60. Die genealogische Stellung des Arabischen in den semi- tischen Sprachen. Ein ungelostes Problem der Semitisk, in: W. Diem - S. Wild (ed.), Studien aus Arabistik und Semitistik. Anton Spitaler zum siebzigsten Geburtstag von seinen Schulern iiberreicht (Wiesbaden) 65-85. Ethiopic Grammar, 2nd ed. by C. Bezold. Trans. J. A. Crichton. London. B. W. W. "Eblaite" = The Earliest Known Dialect of Akkadian. ZDMG 138,211-235. J. A. Knudtzon, Die El-Amarna-Tafeln mit Einleitung und Erlduterungen. 2 vol. Leipzig, 1907-15. Cited by text number. Genetic Subgroupings of the Semitic Languages. PhD Dissertation, University of Texas at Austin. Akkadian Evidence for Proto-Semitic Affricates. ICS 37, 101-107. Interpretation of Orthographic Forms, in: P. Baldi (ed.), Linguistic Change and Reconstruction Methodology (Trends in Linguistics, Studies and Monographs 45; Ber- lin/New York) 619-637. The Diachronic Relationship between Negative and Inter- rogative Markers in Semitic, in: Kaye (1991) 411-429. Genetic Subgrouping of the Semitic Languages, in: Hetz- ron(1997) 3-15. A Grammar of Classical Arabic, 3rd ed. Trans. J. Rod- gers. New Haven. Traite de philologie arabe. 2 vol. Beirut. 196 JOHN HUEHNERGARD Folmer, M. 1995 The Aramaic Language in the Achaemenid Period: A Study in Linguistic Variation. OLA 68. Leuven. Fronzaroli, P. 1981 La congiunzione eblaita ap. Studi Eblaiti 4, 167-176. Fronzaroli, P. (ed.) 1984 Studies on the Language ofEbla. Quaderni di Semitistica 13. Florence. GAG W. von Soden, in collaboration with W. R. Mayer, Grun- driss der akkadischen Grammatik, 3rd ed. = Analecta Orientalia 33, Rome, 1995. Garr, W. R. 1985 Dialect Geography of Syria-Palestine, 1000-586, B.C.E. Philadelphia. Gelb, I. J. 1961 Old Akkadian Writing and Grammar, 2nd ed. MAD 2. Chicago. Gensler, O. D. 1997 Reconstructing Quadriliteral Verb Inflection: Ethiopic, Akkadian, Proto-Semitic. JSSA2, 229-257. Ginsberg, H. L. 1932-33 On the Epic of cAl>eyan Bacl. Tarbiz 4, 106-109, 380-390 (Hebrew). 1939 Two Religious Borrowings in Ugaritic Literature. Orien- talia 8,317-327'. GKC E. Kautzsch, Gesenius' Hebrew Grammar^. Trans. A. E. Cowley, 2nd ed. Oxford, 1910. Goldenberg, G. 1977 The Semitic Languages of Ethiopia and their Classifica- tion. BSOAS 40,461-507. Gordon, C. W. 1991 Qstul Nouns in Classical Hebrew. Abr-Nahrain 29, 83- 86. FEATURES OF CENTRAL SEMITIC 197 GVG Hasselbach, R. 2004 Hayes, J. 1994 Hecker, K. 1984 Hetzron, R. 1973-74 C. Brockelmann, Grundriss der vergleichenden Gram- matik der semitischen Sprachen. 2 vol. Berlin, 1908— 1913. The Markers of Person, Gender and Number in the Pre- fixes of G-Preformative Conjugations in Semitic. JAOS 124,23-35 Traces of Barth's Law in Epigraphic South Arabic. ZDMG 144, 250-258. Doppelt ^-erweiterte Formen oder: Der eblaitische Infini- tiv, in: Cagni (1984) 205-223. The Vocalization of Prefixes in Semitic Active and Pas- sive Verbs. MUSJ4S (FS H. Fleisch, Beirut), 35^18. 1974 La division des langues semitiques, in: A. Caquot - D. Cohen (ed.), Actes du premier Congres international de linguistique semitique et chamito-semitique, Paris 16-19 juillet 1969 (The Hague/Paris) 181-194. 1975 La vocalisation des prefixes verbaux en semitique, in: A. Caquot (ed.), Actes du XXH& Congres international des orientalistes (Paris) 23-24. 1976 Two Principles of Genetic Reconstruction. Lingua 38, 89-108. 1977 The Gunnan-Gurage Languages. Naples. Hetzron, R. (ed.) 1997 The Semitic Languages. London. Holmstedt, R. D. 2000 The Phonology of Classical Hebrew: A Linguistic Study of Long Vowels and Syllable Structure. ZAH13, 145-56. Hommel, F. 1883. Die semitischen Volker und Sprachen als erster Ver- such einer Encyclopadie der semitischen Sprach- und Al- terthums-Wissenschaft. Erster Band: Allgemeine Einlei- tung (Die Bedeutung der Semiten fur die Kultur- geschichte). Erstes Buch: Die vorsemitischen Kulturen in Aegypten und Babylonien. Leipzig. 198 JOHN HUEHNERGARD Hopkins, S. 1984 Studies in the Grammar of Early Arabic. London Oriental Series 37. Oxford. Huehnergard, J. 1987a Northwest Semitic Vocabulary in Akkadian Texts. JAOS 107, 713-725. 1987b "Stative," Predicative, Pseudo-Verb. JNES 46, 215-232. 1987c The Feminine Plural Jussive in Old Aramaic. ZDMG 137, 266-277. 1987d Three Notes on Akkadian Morphology, in: D. Golomb (ed.), "Working with No Data": Semitic and Egyptian Studies Presented to Thomas O. Lambdin (Winona Lake, Ind.) 181-194. 1992 Historical Phonology and the Hebrew Piel, in: W. Bodine (ed.), Linguistics and Biblical Hebrew (Winona Lake, Ind.) 209-229. Izre'el, S. 1991 On the Person Prefixes of the Akkadian Verb. JANES 20, 35-56. Johnstone, T. M. 1970 A Definite Article in the Modern South Arabian Lan- guages. BSOAS 33, 295-307. 1975 The Modern South Arabian Languages. Afroasiatic Lin- guistics 1/5, 1-29 = 93-121. 1977 Harsusi Lexicon and English-Harsusi Word-List. Lon- don. 1981 Jibbali Lexicon. Oxford. 1987 Mehri Lexicon and English-Mehri Word-List. London. Joiion, P. 1923 Grammaire de I'hebreu biblique. Rome. Kaufman, S. A. 1997 Aramaic, in: Hetzron (1997) 114-130. Kautzsch, E. 1906 Die sogenannten aramaisierenden Formen der Verba V"V im Hebraischen, in: C. Bezold (ed.), Orientalische Studien The- odor Noldeke zum siebzigsten Geburtstag (2. Mdrz 1906) gewidmet von Freunden und Schulern (Gieszen) 771-780. FEATURES OF CENTRAL SEMITIC 199 Kaye, A.S. (ed.) 1991 Semitic Studies in Honor of WolfLeslau on the Occasion of his Eighty-fifth Birthday. 2 vol. Wiesbaden. 1991a Etymology, Etymological Method, Phonological Evolu- tion, and Comparative Semitics: Ge'ez (Classical Ethio- pic) ?9gr and Colloquial Syro-Palestinian Arabic ?9zr 'Foot' One Last Time, in Kaye (1991) 826-849. Kienast, B. 1984 Nomina mit T-Prafix und T-Infix in der Sprache von Ebla und ihre sumerischen Aquivalente, in: Cagni (1984) 225- 255. 2001 Historische semitische Sprachwissenschaft. Wiesbaden. Knibb, M. A. 1978 The Ethiopic Book of Enoch: I. Text and Apparatus. Ox- ford. Knudsen, E. E. 1998 Central Semitic *yaqtulum Reconsidered: A Rejoinder to J.Tropper.J5'1S43, 1-9. Kogan, L. E. and A. V. Korotayev 1997 Sayhadic (Epigraphic South Arabian), in: Hetzron (1997) 220-241. Krebernik, M. 1984 Verbalnomina mit pra- und infigiertem t in Ebla. Studi Eblaitil, 191-211. 1988 Die Personennamen der Ebla-Texte: Eine Zwischenbi- lanz. Berliner Beitrage zum Vorderen Orient 7. Berlin. 1996 The Linguistic Classification of Eblaite: Methods, Prob- lems, and Results, in: J. S. Cooper - G. M. Schwartz (ed.), The Study of the Ancient Near East in the Twenty- First Century: The William Foxwell Albright Centennial Conference (Winona Lake, Ind.) 233-249. Lambdin, T. O. 1971 The Junctural Origin of the West Semitic Definite Article, in: H. Goedicke (ed.), Near Eastern Studies in Honor of William Foxwell Albright (Baltimore), 315-333. 1978 Introduction to Classical Ethiopic (Ge'ez). Harvard Se- mitic Studies 24. Missoula, Mont. 200 JOHN HUEHNERGARD Larsen, M. T. 1988 Meillet, A. 1925 Moran, W. L. 2003 Moscati, S. 1957 Miiller, H.-P. 1986 Muraoka, T. 1985 Old Assyrian Texts, in: I. Spar (ed.), Tablets, Cones, and Brickes of the Third and Second Millennia B.C. (Cunei- form Texts in the Metropolitan Museum of Art 1; New York) 92-143. La methode comparative en linguistique historique. Oslo/Paris. Amarna Studies, edited by J. Huehnergard and S. Izre'el. Harvard Semitic Studies 54. Winona Lake, Ind. Plurali interni in ugaritico? RSO 32, 339-352. Aramaisierende Bildungen bei Verba mediae gemina- tae—ein Irrtum der Hebraistik. FT 36, 423^137. Emphatic Words and Structures in Biblical Hebrew. Jeru- salem/Leiden. Muraoka, T. and B. Porten 1998 Nebes, N. 1994 A Grammar of Egyptian Aramaic. Leiden. Zur Form der Imperfektbasis des unvermehrten Grund- stammes im Altsudarabischen, in: W. Heinrichs - G. Schoeler (ed.), Festschrift Ewald Wagner zum 65. Ge- burtstag: I. Semitische Studien unter besonderer Beriick- sichtigung der Sudsemitistik (Beirut/Stuttgart) 59-81. Nebes, N. and P. Stein 2004 Nielsen, H. F. 1989 Noldeke, T. 1899 1904 Epigraphic South Arabian, in: R. Woodard (ed.), The Cambridge Encyclopedia of the World's Ancient Lan- guages (Cambridge) 454-487. The Germanic Languages: Origins and Early Dialectal Interrelations. Tuscaloosa/London. Die semitischen Sprachen. Eine Skizze, 2nd ed. Leipzig. Compendious Syriac Grammar. Trans. J. A. Crichton. London. FEATURES OF CENTRAL SEMITIC 201 Olmo Lete, G. del and J. Sanmartm 1996-2000 Diccionario de la lengua ugaritica. 2 vol. Aula Orientlis Supplementa 7, 8. Barcelona. Porkhomovsky, V. 1997 Modern South Arabian Languages from a Semitic and Hamito-Semitic Perspective. Proceedings of the Seminar for Arabian Studies 27, 219-223. Powell, M. A. 1979 Notes on Akkadian Numbers and Number Syntax. JSS 24, 13-18. Rabin, C. 1951 Ancient West-Arabian. London. Rainey, A. F. 1978 The Barth-Ginsberg Law in the Amarna Tablets. Eretz- IsraelS, 8*-13*. 1996 Canaanite in the Amarna Tablets. 4 vol. Leiden. Ratcliffe, R. R. 1998a Defining Morphological Isoglosses: The "Broken" Plural and Semitic Subclassification. JNES 57, 81-123. 1998b The Broken Plural Problem in Arabic and Comparative Semitic: Allomorphy and Analogy in Non-concatenative Morphology. CILT 168. Amsterdam. Reckendorf, H. 1894 Weitere Duale im Aethiopischen. ZDMG 48, 380. Ricks, S. D. 1989 Lexicon of Inscriptional Qatabanian. Studia Pohl 14. Rome. Rodgers, J. 1991 The Subgrouping of the South Semitic Languages, in: Kaye(1991)1323-1336. Ryckmans, J., W. W. Miiller, Y. M. Abdallah 1994 Textes du Yemen antique inscrits sur bois. Publications de rinstitut Orientaliste de Louvain 43. Louvain. Schall, A. 1982 Geschichte des arabischen Wortschatzes — Lehn- und Fremdworter im Klassischen Arabisch, in: W. Fischer 202 JOHN HUEHNERGARD Schub, M. B. 1973 (ed.), Grundrifi der Arabischen Philologie: I. Sprachwis- senschaft (Wiesbaden) 142-164. A Note on the Dialect of AbT TamTm and Barth's Law. Arabica 20, 307-308 = ZDMG 124 (1974), 13-14 = JNSL 4(1975), 71-73. Simeone-Sennelle, M.-C. 1997 The Modern South Arabian Languages, in: Hetzron (1997) 378-423. Soden, W. von 1961 1983 1984 Streck, M. P. 2000 Testen, D. 1992 1998 Tropper, J. 1997 2000 2001 Verreet, E. 1984 Die Zahlen 20-90 im Semitischen und der Status absolu- tes. WZKM51, 24-28. Zu den semitischen und akkadischen Kardinalzahlen und ihrer Konstruktion. ZA 73, 82-91. Sprachfamilien und Einzelsprachen im Altsemitischen: Akkadisch und Eblaitisch, in: Fronzaroli (1984) 11-24. Das amurritische Onomastikon der altbabylonischen Zeit: I. Die Amurriter, Die onomastische Forschung, Ortho- graphie und Phonologie, Nominalmorphologie. AOAT 271/1. Munster. A Trace of Barth's Preradical *i in Akkadian. JNES 51, 131-133. Parallels in Semitic Linguistics: The Development of Arabic la- and Related Semitic Particles. Studies in Se- mitic Languages and Linguistics 26. Leiden. Subvarianten und Funktionen der sabaischen Prafixkon- jugation. Or. 66, 34-57. Grammatikdes Ugaritischen. AOAT 273. Munster. Die Herausbildung des bestimmten Artikels im Semi- tischen. JSS 46, 1-31. Beobachtungen zum ugaritischen Verbalsystem. Ugarit- Forschungen 16, 307-321. FEATURES OF CENTRAL SEMITIC 203 Voigt, R. M. 1987a 1987b 1988 1994 1995 1998 Die Personalpronomina der 3. Personen im Semitischen. WdO 18,49-63. The Classification of Central Semitic. JSS 32, 1-21. Die infirmen Verbaltypen des Arabischen und das Biradi- kalismus-Problem. Wiesbaden/Stuttgart. Der Lautwandel sl > h in wurzellosen Morphemen des Alt- und Neusildarabischen, in: G. Goldenberg - S. Raz (ed.), Semitic and Cushitic Studies (Wiesbaden) 19-28. Akkadisch summa 'wenn' und die Konditionalpartikeln des Westsemitischen, in: M. Dietrich - O. Loretz (ed.), Vom Alten Orient Zum Alten Testament: Festschrift fur Wolfram Freiherrn von Soden zum 85. Geburtstag am 19. Juni 1993 (AOAT 240; Kevelaer/Neukirchen-Vluyn) 517-528. Der Artikel im Semitischen. JSS 43, 221-258. Waltke, B. and M. O'Connor 1990 Introduction to Biblical Hebrew Syntax. Winona Lake, Ind. Whiting, R. M. 1987 Zaborski, A. 1991 2000 Zemanek, P. 1996 Zewi, T. 2000 Old Babylonian Letters from Tell Asmar. Assyriological Studies 22. Chicago. The Position of Arabic within the Semitic Dialect Contin- uum, in: K. Devenyi - T. Ivanyi (ed.), Proceedings of the Colloquium on Arabic Grammar, Budapest, 1-7 Septem- ber 1991 = The Arabist: Budapest Studies in Arabic 3-4 (Budapest) 365-375. Inflected Article in Proto-Arabic and Some Other West Semitic Languages. Asian and African Studies 9, 24-35. The Origins of Pharyngealization in Semitic. Praha. Is There a Tripartite Nominal Sentence in Biblical He- brew? JNSL 26, 51-63.