Demonstratives in Semitic.
Semitic languages employ a wide variety of demonstrative pronouns to express basic deictic categories. These pronouns commonly reflect a bipartite system that distinguishes between near and remote deixis. Most West Semitic languages use a reflex of a basic element *[delta]V: for near deixis in the singular, as, for example, Hebrew ze (ms) < *[delta]i, Biblical Aramaic da(') and Ge'ez za (fs) < *[delta]a, while far deixis is either expressed by a suffix -k or the anaphoric pronoun, as in BA dek (ms) and Ge'ez zeku (ms) < *[delta]ik(u), Hebrew hu(') and Old South Arabian h' / hw' (3ms anaphoric pronoun). (1) WS near demonstratives in the plural are most commonly construed around a base 'Vl(lV), as in Hebrew 'elle, Old Aramaic 'l, BA 'elle and 'el, while far deixis is, again, expressed by suffixed -k or the respective anaphoric pronoun. Akkadian, the only East Semitic language for which we have sufficient evidence for the deictic system, differs from most classical WS languages in that it only uses one base for all forms of the pronoun expressing near deixis and a second base for far deixis, both of which are fully declined for gender, number, and case. Neither the pronoun for near nor that for far deixis in Akkadian reflects the common WS demonstrative base *[delta]V:.
In addition to these basic demonstratives, Semitic languages have a variety of elements that can be affixed to the pronouns, such as * ha, *n and *l(V), as in Classical Arabic ha[delta]a (ms), [delta]alika (ms far deixis), BA dna (ms), and JPA haden (ms). (2) The affixation of such elements can even lead to the loss of the original demonstrative base, as in Syriac hana (ms), JPA han (ms), which must be derived from original * hadena. (3)
The great variety of elements that can be affixed to each base in individual Semitic languages has led to the reconstruction of a large number of 'demonstratives,' which are only rarely distinguished with regard to function and syntactic context. The absence of a detailed functional analysis is primarily the result of the assumed derivation of these morphemes. Traditional studies of Semitic demonstratives--most of which come from the early twentieth century--derive demonstrative elements from interjections, such as ha and han, which are considered independent elements and, more importantly, share the same basic interjective function. (4)
The most important of these early-twentieth century studies of demonstrative pronouns were undertaken by Barth in 1907 and 1913. Barth assumed that there existed several demonstrative bases--representing 'real' demonstratives--such as annu, z (d), and hahu with the respective plural 'ullu, and a variety of particles that were used to strengthen deixis, including ha, 'a, hei / hai, la, n, tu, ti, and ka. (5) Most of these strengthening elements are considered PS and are, as mentioned above, derived from particles functioning as interjections attested in individual Semitic languages. Only a few of these elements are thought to be secondary derivations. The element 'a, for example, which has the same distribution and function as ha, is derived from ha. (6) Furthermore, Barth assumes that the demonstrative *hai developed into *ai in some languages on the basis of modern Arabic dialects such as Syrian-Arabic haida (ms), haidi (fs), and Aramaic forms such as the expression 'ydy w'ydy 'this and that' attested in the Babylonian Talmud and demonstratives of the shape 'ydk Paydek/. (7) The base *hai itself is considered PS. The reconstruction of this element for PS is doubtful, however, since in those modern Arabic dialects that have hay, this element is secondarily derived from the commonly used demonstrative base ha- and the feminine ending -i, resulting in ha-y. The original feminine form was then extended to the masculine. (8) A similar development underlies the forms in Aramaic, where hay is used for the fem. in various dialects, such as JPA and Syriac, which then was reanalyzed as a general demonstrative base.
The element la is found in Hebrew hallaze and Arabic relative pronouns of the type 'alla[delta]i. According to Barth, la sometimes dissimilated to li, resulting in forms such as [delta]alika (ms). (9) On comparative grounds, it is more likely that *li is the original vocalization of this particle (see section IV below), while la is secondarily derived, perhaps from a combination of li and *ha([+ or -]n). The element n, including its biform -na, is reflected in forms such as Ge'ez zentu (ms) and Aramaic 'illen (cp) and dna (ms). (10) Both n and li will be discussed in more detail below.
Furthermore, Barth suggested that there existed an original demonstrative containing /t/--tu for the masc, ti for the fem., and ta as neuter--that is reflected in Ge'ez zentu (ms), Arabic tilka (fs), and Hebrew zo(')t (fs). (11) Other scholars have connected this /t/ to the fem. ending, not to an original demonstrative element. (12) Although it is true that some AA languages make use of demonstratives with an element /t/, it is disputable whether the /t/ found in Semitic languages reflects an original demonstrative. (13) In Ge'ez, the /t/ of the demonstrative pronouns, which also have biforms without -tu/-ti, is most likely taken from the independent pronouns we'etu (3ms) and ye'eti (3fs). The /t/ in these pronouns represents an original oblique form originally marked by t-insertion, which is also known from other Semitic languages. (14) In other languages, demonstratives with /t/ are primarily used for the fern., as in Hebrew, OSA, Arabic, and Old Aramaic, and should therefore be interpreted as reflecting the fern. ending -t. There is, consequently, no convincing evidence for a reconstruction of a PS demonstrative element *t(V).
The element -ka is the only element that consistently has been assigned a function. It regularly marks far deixis in those languages in which it occurs.
In addition to the elements suggested by Barth, other scholars such as Brockelmann reconstructed a demonstrative m, which presumably underlies the Assyrian far demonstrative ammium and the expansion particle in Mehri -(e)meh. (15) Alternatively, this element has been derived from n and been compared to mimation / nunation attested on nouns. (16)
The most frequently used 'real' demonstrative base, according to Barth, is z / d, which can be extended by all the aforementioned particles. (17) There is no doubt that we can reconstruct a base *[delta]V for WS: that is primarily used for near demonstratives in the singular, but the reconstruction of the original vowel distribution is problematic. Barth assumed that the proto-form, which he considers to be PS, had short /e/ in the masc. and long /a/ in the fern. Arabic subsequently changed the vowel qualities to /a/ in the masc. and /i/ in the fern. on the basis of the fs demonstrative ti. (18) However, the reconstruction of the various vowel qualities as dependent on gender is not generally accepted, as exemplified by the following quote from Fischer (1959):
Es soll damit nicht gesagt werden, da[beta] da, di, du keinerlei Unterschied der Bedeutungsnuance aufwiesen. Jedoch konnte in den bekannten semitischen Sprachen bis jetzt kein Hinweis darauf entdeckt werden, welcher Art die Bedeutungsdifferenzierung gewesen sein konnte. Sicher scheint nur, da[beta] sie nicht Genus und Numerus betraf. (19)
Other explanations for the distribution and function of the original vowels, such as the one of Brockelmann, who, based on Hebrew, suggested that *[delta]l originally represented the genitive from which [delta]u was derived analogically as nominative, have equally not gained wider acceptance.(20) The original distribution and function of the vowels in the base *[delta]V: is still a matter of dispute today. J. Huehnergard, in a recent article about the relative pronouns se- and 'aser in Hebrew, briefly discusses the relationship between the WS det.-rel. pronoun derived from *[delta] and demonstrative pronouns derived from the same base. He states:
The precise nature of the relationship of the relative pronouns to the demonstratives is not entirely clear, nor is the detailed reconstruction of the paradigms of these demonstratives back to Proto-West Semitic. What does seem likely, both on internal grounds and on the basis of cross-linguistic typology, is that the relative pronouns are derived from the demonstratives. (21)
The reconstruction of the common plural base for near deixis is equally uncertain. Barth assumed biforms with different vowel qualities for PS, *'ull and 'ill, while, for example, Bauer and Leander reconstructed forms with only one PS vowel, *i, but with two alternate word-final syllables, *'illa and *'ilay.(22)
An altogether different reconstruction has recently been postulated by Lipinski, who suggests that demonstratives of the basic shape *hanni-, *halli- and *'ulli- should all be derived from one single proto-form, *hanni-, since the alternation between liquids is allegedly a well-attested phenomenon within Semitic. The Assyrian ammium and other m-elements would similarly reflect a variant of * hanni-. (23) This derivation is highly problematic, since, although a certain variation between liquids does indeed occur in Semitic, it is by far not common enough to account for all the varying forms by a simple interchange of these phonemes.
In this article, a new attempt shall be made to reconstruct the most basic demonstrative pronouns and particles that can be affixed to demonstrative bases with special emphasis on functional distinctions of these affixes. The study will further consider a cross-linguistic typological analysis of demonstratives which will be compared to Semitic in order to determine whether this kind of analysis is helpful for the reconstruction and understanding of the Semitic data.
2. GENERAL TYPOLOGY OF DEMONSTRATIVES
The following typological description is based on a detailed study of demonstratives in a large variety of languages undertaken by H. Diessel. (24) The most important issues when considering a typology of demonstratives are, first, agreement hierarchies, that is, how demonstratives in different syntactic positions and functions behave concerning the marking of gender, number, and case in comparison with the normative nominal inflection of the respective language, and, second, the grammaticalization of various types of demonstratives.
Diessel distinguishes four different sub-categories of demonstratives: 1. pronominal demonstratives, which are independent pronouns in the argument position of a verb ('this is the house'); 2. adnominal demonstratives, which accompany a co-occurring noun ('this house')--although it should be noted that only a few languages distinguish categories one and two morphologically; 3. adverbial demonstratives that constitute adverbs of location ('here / there'); and 4. identificational demonstratives, which are pronouns used in copular and nonverbal clauses. (25) Although only the first two categories are of immediate importance for our analysis, adverbial demonstratives will nevertheless be included in this section to exemplify certain hierarchy chains.
Generally, when a language exhibits nominal inflection, pronominal demonstratives are more likely to be inflected than adnominal or adverbial demonstratives, thus
noun > pronominal > adnominal > adverbial
The most common inflectional category of pronominal demonstratives is number, followed by gender and case, leading to the hierarchy
number > gender > case (26)
This means, in cases where an adnominal demonstrative is inflected for number and gender, the pronominal demonstratives show at least the same kind of inflection, while the opposite scenario is not necessarily true. An inflected substantive can be accompanied by an uninflected adnominal demonstrative. In languages in which nouns are inflected for gender, number, and case, the pronominal demonstratives are usually marked for the same grammatical categories, while adnominals are more often uninflected. (27) These general inflectional criteria will become important when we consider Semitic demonstratives.
Demonstratives often grammaticalize into other grammatical markers. The functional categories into which demonstratives grammaticalize are dependent on their original syntactic context. In general, pronominal demonstratives grammaticalize into third person pronouns, relative pronouns, possessive markers, and sometimes into complementizers, while adnominal demonstratives tend to grammaticalize into definite articles and noun class markers. (28)
In the following sections I will first provide an overview of the most common types of demonstratives attested in the Semitic languages, consider their syntactic context and inflectional behavior with regard to the normative nominal inflection of the respective language, determine whether the typology of demonstratives as suggested by Diessel can be applied to Semitic and, if it can, investigate whether it is helpful for the reconstruction of these morphemes in a broader Semitic framework.
3. SEMITIC EVIDENCE
It is impossible to list all the demonstrative pronouns attested in Semitic languages in a study of this scope. The following sections provide an overview of the most frequently found forms and components in the major branches of Semitic, that is, Akkadian, Ethiopian, MSA, OSA, ANA, Arabic, Ugaritic, Hebrew, Phoenician, and Aramaic.
Akkadian has one single base for the pronoun expressing near deixis and one base for the remote demonstrative. Both are fully declined for gender, number, and case, as are nouns in general in Babylonian and Assyrian. The near demonstrative consists of the base *hanni, while the original far demonstrative has two different reflexes in Babylonian and Assyrian. In Babylonian, it is derived from a similar base as the plural demonstratives for near deixis in WS languages, *'ulli-, while the Assyrian form has a base containing /m/. (29) The pronouns for far deixis are not frequently used, the anaphoric pronouns being more often employed instead.
The demonstrative annum is primarily an adjective and follows the noun it modifies as do other adjectives in Akkadian. Only the fs can also be used substantivally and even take pronominal suffixes, as in annitka 'this of yours.' (30)
The demonstrative pronouns for near deixis in Ge'ez have various forms, all of which can be derived from a base *[delta]V(:) in the singular--with or without the addition of -(n)tV--and from a base *'i/ullV in the plural. Far deixis is expressed by the anaphoric pronoun or by a second set of demonstratives based on the near demonstratives plus -k in the ms and pl., while the fs uses a different base. The demonstrative pronouns in Ge'ez distinguish two cases as do ordinary nouns, acc. ~ non-acc. Ethiopian languages commonly distinguish a masc. and fem. demonstrative in the plural. The ms and fs have proclitic and independent forms in the singular that are used interchangeably.
The demonstrative pronouns in Ge'ez are used as both adnominal and pronominal demonstratives. The two functions can be differentiated syntactically by the use of a copula, a third person independent pronoun, that is inserted between the demonstrative and the predicate noun when the demonstrative is used as the subject, as in zati ye'eti walatta negus, 'this is the daughter of the king,' as opposed to za-be' sit and zati be'sit 'this woman.' (31)
In Tigre, we find the leveling of one base in the singular and plural for both the near and far demonstrative. Near deixis is expressed by the same base that is used for the plural in Ge'ez, while the pronoun for far deixis has a base with single/1/loh-. As nouns in general in Tigre, demonstratives are not inflected for case.
All these forms are used both substantivally and adjectivally, as in 'ellan 'am'elat 'these days vs. 'ellima qadam 'ella qer' an 'i'ammer 'this one also does not yet know to read.'
In Tigrinya, demonstratives expressing near deixis likewise leveled one base for all forms, in this case *[delta]V with prosthetic /'e/. Far deixis is expressed by the anaphoric pronoun. Similar to Tigre, Tigrinya does not preserve case inflection.
3.3. MSA (34)
In Harsusi we find the same singular and plural bases for near deixis as in most other WS languages, while far deixis is expressed by the same base with the addition of -k. Most MSA languages have biforms with /n/or/m/, as shown in the case of Harsusi below. Like Tigre and Tigrinya, MSA languages do not preserve case inflection in either nouns or demonstrative pronouns. MSA languages do not distinguish a fp and mp demonstrative.
The demonstrative pronouns are used both as adnominals, in which case they agree fully with the noun they modify, and as pronominals. (35)
3.4. OSA (36)
The demonstrative pronouns for near deixis in OSA are based on the common demonstrative base *[delta]V: in the singular and dual. In the plural, they are derived from a base *'Vl(l)V. Neither seems to be declined for case. However, the pronouns expressing remote deixis, for which OSA uses the anaphoric pronouns, distinguish a nom. and an obl. form. It is not clear whether OSA generally preserved case inflection. (37) The following paradigm is based on Sabaic.
The anaphoric pronouns are used both adjectivally and substantivally, while demonstratives for near deixis are primarily attested in adnominal position. Pronominal usage of near demonstratives is rare, as in 'lt 'hgrn 'these are the towns.' (38)
3.5. ANA (39)
The evidence for demonstrative pronouns in ANA is scarce. In Thamudic, only one form is attested in the singular that is used for both the masculine and fem, and reflects the common demonstrative base *[delta]V:. The same base is attested in Dadanitic, although only adverbially after the preposition b-. There is no evidence for far demonstratives.
Near deixis: ms+fs zn (Thamudic) adverbial: b-dh 'in this' (= 'here') (Dadanitic)
Due to the scarcity of attestations it is difficult to say anything about the syntactic use of demonstratives in ANA. The attestations of zn known so far are pronominal, as in zn gnm bn 'bdmt 'this is PN son of PN,' although the absence of attestations of adjectival use could, of course, be accidental. (40)
3.6. Classical Arabic
In Classical Arabic, the base of demonstratives expressing near deixis is the same as in other WS languages, although it is commonly prefixed with the element ha-. Far deixis is expressed by the same demonstrative base, in which case it usually occurs without ha- and is characterized by a suffix -ka instead. We also often find an element la or li, which can be added in both the singular and plural. (41) The demonstrative pronouns are not inflected for case except in the dual, which has a nom. and non-nom. form for both the masculine and feminine. This inflectional behavior is in sharp contrast to the general noun inflection that distinguishes gender, number, and case in the singular, dual, and plural.
Besides the more common longer forms Classical Arabic has biforms without the prefix ha- to express near deixis. For the fem., there also exist short forms with initial /t/, as found in the fs far demonstrative, although their attestations are rare. In these forms, both /a/ and /i/ are used for the feminine, as opposed to the otherwise typical distribution of /a/ for the masc. and /i/ for the fem. (44)
The demonstratives are used both adjectivally and substantivally. In both cases they precede the noun they modify. (45) When the demonstrative is used as subject together with a definite predicate, a personal pronoun is inserted between demonstrative and nominal predicate in order to distinguish this construction from the attributive use, as in ha[delta]a rrajulu 'this man' vs. ha[delta]a huwa rrajulu 'this is the man.'
3.7. Modern Arabic Dialects
Modern Arabic dialects employ a wide variety of demonstrative pronouns. It is not possible to list all forms in the present study; thus only a few representative examples will be quoted.
Some modern Arabic dialects preserve a very basic variant of demonstrative pronouns, similar to the basic forms attested in Old Arabic. These forms are primarily found in northern African dialects, for which Moroccan has been chosen as a representative, and in dialects from South Arabia. The more frequently found forms are compounds with ha-, which are attested in basically every modern Arabic dialect. The particle ha- is the most basic form of demonstratives for near deixis in many dialects, although it is, by itself, only used attributively, for both the ms and fs, and precedes the noun it modifies, as in ha-qalbek (Tunisian) 'this heart of yours.' (46) The demonstrative particle ha- can be declined for gender and number by the addition of -i in the fern, and -u for the plural, that is, *ha-i and *ha-u, which subsequently contracted to hay and haw. Forms like these are found in the dialects of Syria, Lebanon, and Iraq. (47) Modern Arabic dialects do not usually employ the suffix -n that is found in OSA, MSA, and Aramaic, the only exception being Maltese, which has forms with and without -n. (48) For the plural, we find various forms which sometimes still reflect the WS plural base *VI(l)V, and sometimes analogical derivations on the basis of the common plural marker -u, such as [delta]u, and haw < *ha-u.(49)
The demonstrative pronouns are used both attributively and substantvially, as in ha i lulaja 'this village' vs. ha i lli riditha 'this is what I wanted' (Iraq), Imra ha[delta]i 'this woman' vs. ha[delta]i hey 'this is she' (Dofar). (52) Syntactically, demonstrative pronouns are more flexible in modern Arabic dialects than in the formal language in that they can usually either precede or follow the noun they modify. (53)
Demonstrative pronouns are only rarely attested in Ugaritic and their analysis is far from being certain. The forms that can be analyzed as demonstratives with more or less certainty are compounds consisting of the deictic element hVnnV + certain particles, d for near deixis and k for far deixis. Thus, they are derived from the same base as the Akkadian demonstrative annum. The affix /t/ can be suffixed to the basic form, although the distribution and function of the t-containing forms is not entirely certain. They were not used to distinguish gender or case. (54)
The form hnd is used as a demonstrative for both genders and numbers, always in adjectival function, as in mspr hnd 'this story' and spr mlk hnd 'this document of the king,' while hndt is primarily used as a local adverb 'here.' (55) There seems to be a plural and possibly dual demonstrative in the oblique case hnhmt, a compound of the demonstrative particle * hann and the 3mp oblique independent pronoun, which is likewise only used adjectivally. The form hnk is used as local adverb 'there' and as demonstrative indicating far deixis, while the function of hnkt is still unclear. (56)
The demonstratives for near deixis are based on the element * [delta]V: in the singular and *'illV in the plural. Neither is declined for case, in accordance with the common loss of case marking in Hebrew. The plural also lacks gender distinction. Far deixis is exclusively expressed by the anaphoric pronoun and, contrary to the demonstratives expressing near deixis, distinguishes a masc. and fem, form in the plural. When the demonstrative pronouns are used adjectivally, they take the definite article in accordance with the rules applying to other adjectives, as in ze ha-'is 'this is the man' vs. ha-'is haz-ze 'this man.'
It is disputed whether Hebrew has a demonstrative zu, used for both genders and numbers, or whether this form is only used as a relative pronoun. (59)
Furthermore, Hebrew has biforms for expressing near deixis in the singular that are composed of the above-mentioned demonstrative base plus a prefix halla-, that is, the definite article plus an original element *la-. These forms are rare in Biblical Hebrew, but become more frequent in post-Biblical, especially Rabbinic Hebrew. (60)
3.10. Phoenician (61)
Only the Phoenician forms will be listed here since the Punic demonstratives can be directly derived from Phoenician. Near demonstratives use the same base as Hebrew, while far deixis is expressed by the anaphoric pronoun. There is no distinction of case in the demonstrative pronouns as far as we can tell. In general, Phoenician seems to have lost its case inflection at the beginning of the first millennium B.C.E., although vestiges are still found, for example, before pronominal suffixes, which distinguish a nom./acc. form vs. a genitive. (62)
The plural of the near deictic demonstrative was most likely vocalized with /i/ in the first syllable, based on attestations from the Punic variant found in Poenulus, illii and ily. (65)
Most of the attestations of the demonstrative pronouns are adjectival, as in b'rn zn 'in this coffin.' Friedrich, Rollig, Amadasi-Guzzo have suggested that in cases where an inscription has more than one form of a demonstrative pronoun, as in the inscription of Yehaumilk, which uses both zn and z, these forms might be used to distinguish different gradations of distance, one meaning 'this ... here,' while the other expresses 'that ... over there.' (66)
Similar to Modern Arabic dialects, Aramaic employs various forms of demonstrative pronouns in different dialects. Only some are chosen below to provide a representative sample of the evidence. No attempt has been made to vocalize forms that are only attested in unvocalized scripts.
All these demonstratives can be used both substantivally and adjectivally, as, for example, Egyptian Aramaic sprh znh and 'grt' z' 'this letter' vs. dnh thwmwhy 'this is its boundaries.' (71)
In the Palestinian Talmud, den, dna, da(') are only used substantivally, while all other demonstrative forms are used both as adjectives and nouns. In Targum Onqelos, den, da('), and 'illen are used substantivally, while demonstratives with prefixed ha- are used adjectivally, as in den malkana 'this is our king' vs. malka haden 'this king.' (74)
In Syriac, all demonstratives can be used as substantives and adjectives. (75)
3.12. Summary of the Evidence
We can distinguish several types of demonstrative formations in Semitic. First, a simple formation in the singular based on the element *[delta]V: that is attested with all three PS vowel qualities, *[delta]a, *[delta]i, and *[delta]u, although to different degrees. Languages that make use of these forms include Ge'ez, MSA, ANA, Arabic, Hebrew, Phoenician, and Aramaic. That is, this basic form is attested in all sub-branches of WS. Akkadian/East Semitic is an exception to this distribution since there is no evidence for this base with demonstrative function.
The distribution of vowel quality and quantity in this element differs from language to language. The vowels are never used to distinguish case. They are primarily employed to distinguish gender. The form *[delta]i is attested for the ms demonstrative in Ge'ez, Hebrew, and Aramaic, all of which seem to reflect an original short vowel. (76) The basic form *[delta] a, with original long vowel, is used for the fem. in the same languages, in addition, perhaps, to Phoenician.
The opposite distribution, *[delta]a for the masc. and *[delta]i for the fem.--note that both variants primarily reflect original long vowels--is attested in MSA and Arabic. Since MSA preserves the same basic forms of the demonstratives as some of the southern Arabian Arabic dialects, we might be dealing with an areal phenomenon in this case.
The form *[delta]u is only attested in a few modern Arabic dialects (see section 3.7 above), where it is used for the plural. In these dialects, [delta]u has been secondarily created by analogy with the singular forms and the general plural ending -u. (77) The absence of *[sigma]u in other WS languages is not surprising, since the cp is usually expressed by a different base.
The demonstrative base [delta]V: can be extended by various particles. A very common prefix is the particle *ha-, used in Classical Arabic, modern Arabic dialects, and in various forms of Aramaic, including JPA, CPA, the Aramaic of Targum Onqelos, and Syriac, that is, mostly Middle and Late Aramaic dialects. A sub-variant of *ha- is the particle *han- < *ha +n, which is used in Hebrew in forms such as hallaze, where it reflects the Hebrew definite article, (78) and as the base for the near demonstrative in Akkadian and Ugaritic.
The second common way to expand the demonstrative base [delta]V: is by the suffix -n. This suffix is found in Ge'ez, MSA, OSA, ANA, Maltese, Byblian Phoenician, and in those Aramaic dialects that do not use the prefix ha- extensively, such as Old, Egyptian, BA, and Syriac, although a certain overlap between the use of the prefix ha- and the suffix -n does exist in Aramaic. Interestingly, other languages that have forms with -n rarely use the prefix ha-, Maltese being an exception. The vowel of the base [delta]V: is regularly shortened when the suffix -n is attached, as in Ge'ez zentu * < *zi-n-tu and JPA den (written plene as den) < *din.
A less frequent variant of the suffix -n is the longer form *-na, which is attested in Maltese, where forms both with and without final -a occur, and in Aramaic, particularly in Old, Egyptian, Biblical Aramaic, and Syriac. The demonstratives extended with -n / -na are sometimes exclusively used for the ms (Ge'ez, OSA, Aramaic), and sometimes for both the ms and fs (MSA, ANA, Maltese).
The last element that can be suffixed to singular demonstratives of the base *[delta]V: to be mentioned in this context is the ending -t(V:). This suffix is found in Ge'ez for both the ms and fs--where it probably reflects an analogical extension of the endings of the anaphoric pronouns--in OSA, Hebrew, and Phoenician. In the latter languages it is only used for the fs. (79)
The near demonstrative in the plural reflects the same base in most WS languages, *VllV, which is also attested in Akkadian, where it is used as the far demonstrative. In Ge'ez, the final vowels are secondary and reflect the masc. and fem. plural markers of the verbal inflection, while Tigre leveled this base for all near demonstratives. The same base is used in Hebrew and Aramaic and probably underlies the forms in OSA. The vowel of the first syllable varies between *i and *u. Akkadian has *u, while Hebrew and Aramaic have *i. No evidence for the original vowel quality can be gathered from OSA or Phoenician. Similarly, the shwa in Ge'ez can be derived from either *i or *u.
The plural base has a shorter variant, *'Vl, attested in MSA, Arabic, Hebrew, and Aramaic. Again, the vowel of the first syllable varies. It is *u in Arabic and *i in Hebrew and Aramaic. The MSA form can be derived from either *'il or *'ul.
Both plural bases can be extended in the same way as the singular. The suffix -n is by far the most frequently used extension (Ge'ez, MSA, OSA, Aramaic). A few modern Arabic dialects also use a suffix or infix *li for forms of the plural near demonstrative. However, this element is probably more original to the far demonstrative and will be discussed below.
The demonstratives used for far deixis can be divided into two main categories: 1. demonstratives based on the near demonstrative pronoun plus suffix, usually -k(a); and 2. anaphoric pronouns. The exclusive use of anaphoric pronouns is found in Tigrinya, OSA, Hebrew, and Phoenician. Several languages employ both the anaphoric pronoun and a separate demonstrative. Note, for example, Akkadian, where the use of the anaphoric pronoun is more common than the use of the far demonstrative ullum l ammium, and Ge'ez, where the anaphoric pronoun is also the more frequently used form while demonstratives with -k suffix can have a pejorative connotation. The exclusive use of forms based on the near demonstrative plus suffix is attested in MSA, Arabic, Aramaic, and perhaps Ugaritic. (80) The reduced use of the original demonstrative pronouns in those languages that employ both ways to express remote deixis suggests that the use of the anaphoric pronoun is a later development. This interpretation is based on Kurylowicz' fourth law of analogy, which states that when younger forms take over the function of older ones, and the older ones are preserved, the latter acquire secondary functions. (81)
The most basic form of the remote demonstrative, the base of the near demonstrative plus suffix -k(a), is attested in Ge'ez, MSA, Arabic (particularly in modern Arabic dialects), and in Aramaic, although in the latter it does not occur for all forms of a given paradigm, The ms in Old, Egyptian, and Biblical Aramaic usually does not show the n of the ms demonstrative expressing near deixis. When the remote demonstrative does have n, it is suffixed after the remote deixis indicator -k-. The formations with -k become less frequent in Late Aramaic, where they are replaced by forms consisting of ha- + anaphoric pronoun.
Another suffix primarily found in demonstratives used for far deixis is the element li. This element is found in Arabic, both classical and modern, and can freely be combined with other suffixes such as -k(a).
All Semitic languages discussed in our context distinguish gender (at least in the singular) and number, independent of whether they use a different base for the plural or not. A striking feature is that many languages do not have a distinct fp demonstrative, despite the fact that the nominal and pronominal inflection does distinguish between mp and fp in the respective languages. Languages with a cp include MSA, Arabic, Hebrew, and most Aramaic dialects, Syriac being an exception. The fp demonstrative pronouns attested in Semitic can be explained as secondary developments based on the anaphoric pronouns and/or pronominal suffixes. This secondary derivation is further confirmed by the fact that these pronouns vary widely from language to language and cannot be derived from a common proto-form.
Since many languages, including MSA, Tigre, Tigrinya, Hebrew, and Aramaic, lost case distinction in the nominal inflection, we do not expect demonstratives to be inflected for case in these languages. Interesting for the present discussion are those languages that preserve case inflection, either fully or in part. Akkadian inflects the demonstrative pronouns like every other adjective in strict concordance with the general nominal inflection. Ge'ez has an ace. vs. non-acc. for most demonstrative pronouns, in agreement with the two-case system of the language. OSA, for which the status of case inflection is unclear, clearly preserves a nom. and obl. form of the remote demonstrative expressed by the anaphoric pronoun. Classical Arabic does not inflect demonstrative pronouns in the singular and plural, despite the fact that Arabic nouns and adjectives are fully declined for case, although case distinction is attested in the dual where it is based on the nominal inflection. Thus, case agreement is not as strict as the agreement for gender and number. On the contrary, Arabic clearly lacks the expected case agreement, while OSA only has case agreement in the anaphoric pronoun and not, at least as far as we can tell, in the original demonstrative base.
Most of the demonstrative pronouns can be used both substantivally and adjectivally. This is particularly true for those languages that only have one form for each person and deictic function as well as for those that do not use a definite article. An important exception is Akkadian, in which annum is only used adjectivally. In languages that have more than one form, we can sometimes distinguish different usages among the various formations. Of particular interest are different functions of demonstrative pronouns with and without the prefix ha-. In Arabic dialects the prefix ha- was originally used to strengthen the demonstrative notion. This strengthening function was mostly lost, so that forms with ha- become the normative formations. An interesting distribution of forms with and without ha- is found in the Aramaic of the Palestinian Targums and Targum Onqelos. In the former, short forms without this prefix are used substantivally, while all other forms can be used both as adjectives and substantives. In Targum Onqclos, forms without ha- are used substantivally, while forms with ha- are functionally equivalent to adjectives. A similar situation is attested in Hebrew, if we can count the use of the definite article as a parallel or at least similar usage to that of ha- in Aramaic. (82) As substantives, demonstrative pronouns do not take the definite article, although they do when used adjectivally, that is, ze ha-is 'this is the man,' vs. ha-'is haz-ze 'this man.' (83) In Classical Arabic, demonstratives never take the definite article, independent of their syntactic function. Similarly, demonstratives in Aramaic are not attested in the emphatic state, which equals the definite articles in Hebrew and Arabic. This absence of co-occurrences of the definite article and the particle ha- in Aramaic and Arabic can be explained by the incompatibility of ha- in the function of a demonstrative particle and the def. articles that are equally derived from demonstrative elements--in Aramaic even from the same element that constitutes the base of the definite article in this language (see section 4 below).
4. RECONSTRUCTION OF SEMITIC DEMONSTRATIVES
When we look at the agreement of demonstrative pronouns compared to the nominal inflection in Semitic languages, we notice that demonstratives hardly ever exhibit the same inflectional agreement as nouns and adjectives. This is surprising, given the fact that pronominal demonstratives usually show the same inflection as nouns throughout the world's languages, while adnominals are the demonstrative category that more commonly lacks this agreement. (84) This might suggest at first sight that demonstratives in Semitic are primarily adnominal or derived from adnominals, but this is unlikely for various reasons. First of all, most demonstrative pronouns can be used both substantivally and adjectivally. Furthermore, there is evidence that the relative pronouns of various Semitic languages are derived from original demonstratives, which suggests a derivation from pronominal, not adnominal demonstratives, since grammaticalization of adnominals results, among others, in definite articles but not in relative pronouns. The grammaticalization of adnominal demonstratives is attested in various Semitic languages, although--and this is important to note--in most cases the demonstrative elements underlying definite articles in Semitic do not reflect the same elements as those that grammaticalized into relative pronouns. Furthermore, definite articles developed only in a limited set of Semitic languages and do not, in contrast to relative pronouns, constitute morphemes that can be reconstructed for PS.
Let us first have a look at the relative pronouns that can be connected to demonstratives and then at the particles underlying the definite articles of Semitic languages.
The relative pronouns significant for this investigation are primarily related to the demonstrative base *[delta]V:. In Old Akkadian, and probably also in Eblaite, the relative pronoun has a base *[theta]V: instead of WS *[delta]V:. (85) In both languages it is fully inflected for person, number, and case.
Old Akkadian relative pronouns: (86)
A similar fully declined variant of the base [delta]V: is known from CA, where it does not function as a relative pronoun but indicates possession. The inflection is similar to that found in Old Akkadian and Eblaite in the singular, while the plural, particularly the mp, differs significantly. Also note that CA has biforms in the pl. that reflect a different base.
Ugaritic and OSA have similar paradigms that distinguish between ms, fs, and pl., although the writing without vowels does not allow us to reconstruct the full paradigm in these languages. Ugaritic, like Akkadian, has one base for the singular and plural, while OSA has two bases, *[delta]V: in the singular and *'Vl- in the plural.
The CA relative pronoun also reflects a base [delta]V: in the ms and cp, although it is only used in combination with prefixed 'alla-. The fs has ti, not [delta]V:, while the ms has a final /i/ -vowel, as opposed to the /a/-vowel that is found in the ms demonstrative pronoun. None of these pronouns is inflected for case, except in the dual.
Other Semitic languages that use a reflex of [delta]V: for the relative pronoun equally do not show case distinction, even if the nominal inflection itself does, as, for example, Ge'ez, which has an acc. ~ non-acc. in the nominal inflection but only one form for the relative pronoun, za- (ms).(88) The relative pronoun is inflected for gender and number in Ge'ez and Arabic and for number in MSA. Other languages, such as Aramaic and Hebrew, use a single form for the masc, fern., sing, and pl., that is, the inflectional agreement is more reduced than in the demonstrative pronouns of the respective languages:
The various WS languages reflect all three original vowel qualities; thus we can safely assume that PWS must have had at least three forms of the relative pronoun in the singular. It is likely, based on the evidence from CA in conjunction with the evidence from ES, that WS had a fully declined det.-rel. pronoun. (90) The question remains whether this det.-rel. pronoun had one base for the sing, and pl., or two. Huehnergard assumes that WS plural relative pronouns reflecting *'Vl(lV) are secondarily derived from plural demonstratives and that the plural of the det.-rel. pronoun originally had the same base as the singular--resulting in *[delta]ut for the mp--since it would otherwise be difficult to account for the /t/ in Ugaritic. (91) The plural forms of the Arabic possessive pronoun are likewise considered secondary by Huehnergard, "modeled on the nominal external masculine plural construct endings." (92) Given that the Arabic plurals are most likely secondary and that the rest of the WS languages that have a distinct plural relative pronoun use a base *Vl(lV)--note also that we still find biforms reflecting the p1. base *'ul- in Arabic as well--Huehnergard's reconstruction of the WS plural is not necessarily convincing. Furthermore, the WS det.-rel. pronoun is most likely a grammaticalized form of the demonstrative pronouns, which show no evidence for an original plural base *[delta]V:. Consequently, an original p1. base *'Vl is more likely and supported by typological arguments than an original base *[delta]V:. The Ugaritic/t/in the mp can alternatively be explained as based on an oblique form, similar to/t/-forms attested in the anaphoric pronouns. Thus, the plural relative pronouns with the base * Vl(lV) are the more original forms, while paradigms in languages that have a base *[delta]V: in both the singular and plural are the result of leveling.
In summary, the West and East Semitic relative pronouns reflect the two typical WS demonstrative bases, *[delta]V: and *'Vl(lV), in the same distribution as attested for the demonstratives in WS, namely *[delta]V: in the singular and *'Vl(lV) in the plural. If this analysis is correct, we would expect the demonstratives that constituted the basis for the relative pronouns to be pronominal demonstratives.
The definite articles that developed in various Semitic languages are equally grammaticalized demonstratives, although the original forms underlying the definite articles are heavily disputed. (93) In a recent study, A. Rubin suggested that the CS articles can be derived from two demonstrative particles, *han in the case of Canaanite, Aramaic, ANA, and OSA, and *'ul- in the case of Arabic. (94) That is, Arabic presumably grammaticalized the plural base of the WS demonstrative while the rest of CS used a particle that was not part of a primary demonstrative pronoun.
The demonstrative particle *han is derived from *ha+n. The closing of the syllable by/n/resulted in the shortening of the original long vowel. This particle, although mostly without the addition of /n/, is very frequently attested in combination with demonstrative pronouns, as has been shown in the previous section. It has also been mentioned that there is a tendency for demonstrative pronouns with ha-to be used attributively in those languages that have more than one form. In Akkadian, the base of the near demonstrative is the same *han, and, interestingly, although there is no alternate form, it is primarily used adjectivally. All this points to an interpretation of ha as marking the adnominal use of demonstrative pronouns as opposed to the pronominal use of forms without this element. This functional difference between demonstrative pronouns with and without ha-would fit nicely into the cross-linguistic phenomenon that adnominals, not pronominals, grammaticalize into definite articles in languages that distinguish the two types morphologically. (95) Thus, the demonstrative particle *ha-, including its extended form *han-, was originally used to mark adnominal function. This also conforms to the inflectional behavior of *ha(n). This particle is not generally inflected for gender, number, or case. In those cases in which we do find inflection it is secondary. In languages that do not use the particle ha-, *[delta]V: has both adnominal and pronominal function.
The demonstrative base *'Vl(lV) can, as argued above, likewise function pronominally and adnominally. The adnominal use can be deduced from the Arabic definite article 'al-, which, as Rubin argued, should be derived from the WS plural demonstrative base *'Vl(lV). At the same time, *'Vl(lV) functions in the same way as *[delta]V: in WS, that is, substantially, and can be extended by ha- to mark it as adnominal. The pronominal use underlies the WS demonstrative pronouns and the relative pronouns that grammaticalized out of this base, while the adnominal use underlies the Arabic definite article. The same is true for the WS singular base *[delta]V:. Consequently, both *[delta]V: and *'l(lV) were originally used substantivally and adjectivally. However, already in early Semitic there existed the possibility of marking the demonstrative pronouns as adnominals by affixing the element * ha(n), since the Akkadian near demonstrative reflects the use of this element as adnominal marker.
In sum, both demonstrative bases found in WS languages and partly in ES could originally be used adnominally and pronominally. Early in Semitic, before the split into East and West Semitic, adnominal use could also be marked by specific affixes, particularly *ha and its extended variant * han. Some languages subsequently developed two sets of demonstrative pronouns reflecting the two different syntactic functions, one unmarked and one marked by the affixation of *ha([+ or -]n). This implies that *ha([+ or -]n) was not originally used independently as a demonstrative pronoun but functioned as an adnominal marker. Uses of this element as an independent demonstrative are secondary, as is still obvious in the case of the modern Arabic dialects cited above.
It is likely that other particles that are affixed to demonstrative bases had specific syntactic functions as well, although they are much more difficult to trace on the basis of inner-Semitic evidence. The element -n is often connected with near deixis in Semitic. A demonstrative suffix -n is also known from Egyptian, where it constitutes the most frequently used variant of four different demonstrative suffixes, and from Cushitic and Berber, where it is connected to near deixis. Thus, this element is of ancient AA stock and was probably originally connected to near deixis. (96) In Semitic, -n was replaced by *[delta]V and acquired a secondary function. Since in many languages ha- and -n do not occur together, the affixation of -n might have acquired a similar adnominal marking function as * ha-, although this must remain hypothetical. (97)
The element li likewise still reflects an originally distinct function. It is primarily used in Arabic remote demonstrative pronouns, although it also underlies the second syllable of the long forms of the demonstrative base *'VllV attested in Akkadian, Hebrew, and Aramaic. As mentioned in section 3.6 above, it has been suggested by some Arab grammarians that the pronoun [delta]alika as opposed to [delta]aka expresses the more remote object when two objects are compared. A differentiation of spatial concepts of this kind, that is, with more than just two entities 'this' and 'that,' is not unusual cross-linguistically, and is also attested in other Semitic languages. In Amharic, for example, the deictic gradation of demonstrative pronouns is more complex than the bipartite distribution usually assumed for other Semitic languages. The use of demonstratives in Amharic is dependent on the hearer's and speaker's distance from the object. The demonstrative pronoun ya 'that' (ms) is used when the object is distant from both speaker and hearer, while the anaphoric pronoun essu (ms) is used when an object is close to the hearer but far from the speaker. (98) A base *ll connected to remote deixis is also known from Rendille (Cushitic). (99) Thus, we might have a case similar to that suggested for -n. The demonstrative element *l goes back to AA, where it probably primarily expressed far deixis. In Semitie, it was replaced by other demonstrative elements, notably -ka and the anaphoric pronoun, and, in some cases acquired a secondary function denoting a different degree of remote deixis than -ka, while, in others it lost its original demonstrative force.
Thus, functional differences and/or the substitution of older AA demonstrative elements that were marginalized and acquired secondary meanings in Semitic can account for most of the elements that are affixed to demonstrative bases.
This implies that the most original Semitic demonstratives should be reconstructed without affixes such as *ha-, -n, and -li, since these are functional elements that are not original to the demonstrative pronouns. The most basic pronominal morphemes in Semitic are *[delta]V: and *'Vl. The greatest problem regarding the original reflex of these bases is the distribution of the morpheme-inherent vowels. It still has to be determined whether the singular base should be reconstructed with all three vowels, and if so, how they were functionally distributed, and whether the base-inherent vowel in the plural was /i/ or /u/.
The grammaticalized forms of the base *[delta]V;, reflected in the ES relative pronouns and the Classical Arabic pronouns indicating possession, suggest that the original morpheme had three vowels in the singular that were used to distinguish case. In the typological hierarchy, case is the lowest inflectional category; thus it is not surprising that demonstratives in Semitic are only rarely inflected for case, since case is the first inflectional category to be lost. The only objection to a reconstruction of the singular demonstrative with three proto-forms distinguished by vowel quality that marks case is the fact that no Semitic language preserves this distribution in the original demonstrative morphemes. In those WS languages that have case distinction, it is most likely secondary, as in Ge'ez and the CA dual. It is not even certain that we can reconstruct a demonstrative with original vowel /u/. The most probable reconstruction of the base *5V: that has any foundation on inner-Semitic evidence--primarily WS evidence--is to assume two original vowel qualities, /a/ and /i/, used to distinguish gender: /i/ for the masculine and /a/ for the feminine. The opposite distribution in Arabic is secondary, based on a reinterpretation of /i/ as the feminine marker used in the verbal system and the independent pronouns. The secondary status of the Arabic vowel distribution is confirmed by the fact that the ms relative pronoun in Arabic preserves the original /i/-vowel and that there exists an old fern, demonstrative with /a/, ta.
Thus, at least for PWS, we can reconstruct the following singular forms:
The vowels were shortened when a consonantal suffix was attached, as in *[delta]i+n > *[delta]in.
The plural base in WS had the shape *'Vl in its most basic form, with the element *li being suffixed in various languages. The reconstruction of the morpheme-inherent vowel is, as mentioned above, problematic. The NWS languages for which we have evidence for the vowel quality point to original *i, while Arabic and Akkadian have *u. The evidence from Ethiopian languages is not conclusive since /e/ can be derived from both *i and *u. On the basis of Arabic and Akkadian it is more likely that the original vowel was *u, since these two languages are distant from each other, while *i is attested in closely-related languages in which it could reflect a shared innovation or an areal phenomenon. In Hebrew, Phoenician, and Aramaic, the longer form of the plural base *illi is commonly used, while the shorter form *il is less frequent. It is also in these languages that we find an /i/ vowel in the demonstrative base. The original /u/ from the base *ulli assimilated to the /i/ of the second syllable, since patterns of the shape *quttil do not exist in these languages. (100) The pattern *qittil on the other hand is attested in both the nominal and verbal system, at least in Hebrew, and might have triggered the change of *ulli > *ulli. The shorter form with *i is an analogical formation based on the longer variant. The plural did not originally distinguish gender. As mentioned above, languages that have both a fp and mp developed the distinction secondarily, based on the pronominal and/or nominal inflection, as is evident in Ge'ez, Tigre, OSA, and Syriac. For PWS we can consequently reconstruct the following plural base:
cp *ul([+ or -]li)
Thus, we have a limited set of basic demonstratives in PWS, two forms in the singular in which gender is distinguished by vowel Ablaut, and one plural base. The form *'ulli also underlies the Babylonian far demonstrative ullum. This base should consequently go back to PS. In Babylonian, ullum was leveled and used as remote demonstrative for all persons and numbers. The leveling of this base also occurred in Tigre, where it is used to express near deixis. Note that Ge'ez uses this base for the plural. Furthermore, Assyrian leveled a different base than Babylonian, which is likewise an indication that we are not dealing with the original PS distribution of these forms in Akkadian but have to assume inner-Akkadian changes.
This still leaves us with the PS situation to consider. Since Akkadian underwent internal changes regarding its demonstrative pronouns that caused it to become significantly different from other Semitic languages, it is difficult to say anything certain about the PS reconstruction, and the following observations thus have to remain hypothetical. If the above-suggested derivation of the common Semitic relative pronouns by grammaticalization from demonstratives is correct, the reconstruction of the det.-rel. pronoun should be helpful in determining the original paradigm of the demonstrative pronouns. (101) The case inflection in ES relative and Arabic possessive pronouns should be considered original, since the similarities between these two branches are too great to be coincidental, at least in the singular and dual. Several WS languages confirm the reconstruction of the singular with three vowels, since all three vowel qualities are attested in relative pronouns in this branch of Semitic. The evidence from both ES and WS implies that a fully declined det.-rel. pronoun with at least a sing, base *[delta]V: or *[theta]V: should be reconstructed for PS. Furthermore, the derivation of these pronouns from demonstratives implies that the base *[delta]V: or *[theta]V: has to be reconstructed as a demonstrative base for PS. The question remains whether this base showed the same case inflection as the det.-rel. pronoun, or whether the case inflection of the latter is an innovation based on the nominal inflection. If we follow the typological analysis, it is more likely that the demonstratives were originally declined for all three PS inflectional categories, namely gender, number, and case. (102) The det.-rel. pronoun was grammaticalized from these fully declined demonstratives at an early stage of Semitic, before the separation into ES and WS. Subsequently, since case is the lowest member of the hierarchy chain, it was lost. This loss occurred no later than PWS, although the det.-rel. pronoun survived with case inflection in WS before it was leveled into one variant in those languages that subsequently lost all or part of the original case system. Thus, we can reconstruct the following demonstratives for PS in the singular:
Since the dual is generally based on the singular, it should reflect the same base in the case of the demonstrative pronouns as well:
The plural did not share the same base as the singular and dual, but used *'ul, which is reflected in the WS plural demonstrative and the Akkadian far demonstrative. Whether this base was a comunis form or had separate mp and fp formations is uncertain. Besides Akkadian, only OSA has a separate fp of this base, 'lt vs. masc. 'ln, although this form is probably secondary and based on the gender distinction found in the anaphoric pronoun used as far demonstrative. (103)
These bases could be extended by various affixes, most of which go back to PS and/or AA, such as ha- as marker for adnominal use in PS, -n originally marking near deixis and going back to older stages of AA, -ka as marker for remote deixis in at least PWS (although the morpheme itself goes back to AA), and -li in PS, which probably also originally functioned as marker for far deixis. The individual Semitic languages subsequently underwent various processes of leveling and analogical formation, the latter primarily based on independent pronouns and pronominal suffixes.
In conclusion, the various demonstrative pronouns in Semitic can be reduced to two main demonstrative bases, *[delta]V: as near demonstrative in the singular, which was originally inflected for case, gender, and number; and *'ul in the plural, which showed the same type of inflectional categories as the singular base. These bases could be expanded and semantically modified by several affixes, most of which go back to ancient AA demonstrative elements. However, their number has to be reduced compared to the affixes listed by scholars such as Barth and Brockelmann, since some of the affixes reconstructed previously developed secondarily or are not connected to demonstratives, as has been argued in section 1 above. The affixes that can be reconstructed for PS are *ha-, *-n, and *-li, and probably -ka; the latter's PS status can only be inferred from broader AA evidence.
The most common suffix to denote far deixis is -ka. The suffix -li is less frequent. It is a vestige of an older AA demonstrative element primarily used to mark far deixis, although it lost its demonstrative force in Semitic when its function was taken over by -ka. The prefix ha-, including its biform *han-, originally marked adnominal usage and developed into the definite article in various CS languages. The common suffix -n is of ancient AA stock and can be connected to near deixis, although, similar to the element li, it lost its original demonstrative force and was replaced by other elements in Semitic. Both elements, *-n and *li, are still reflected with their original near and far deictic connotations in the Akkadian demonstratives annum and ullum and probably triggered the leveling of these bases. This implies that, early in Semitic, these two affixes still had distinct deictic functions, although they were not the primary markers of near and far deixis but were replaced by *[delta]V: and *'ul.
A morphological reconstruction can only be suggested for PWS on the basis of inner-Semitic evidence, from which we can assume two forms of the base in the singular, a ms and fs, which are distinguished by different vowel qualities (i.e., *[delta]i in the masc. and *[delta]a in the fem.), while the plural had a comunis form reflecting the base *'ul. Since the reflexes of these demonstrative bases do not show case inflection in most of the Semitic languages, even when nouns and adjectives are fully inflected for case as in Classical Arabic, it is unlikely that PWS demonstratives were inflected for case.
The PS situation can only be deduced by comparison with the behavior of demonstratives in language families other than Semitic, which, necessarily, only leads to a hypothetical reconstruction. Based on general language typology, pronominal demonstratives grammaticalize into relative pronouns and adnominal demonstratives into definite articles. This is exactly what we find in Semitic. The demonstrative bases reconstructed above were grammaticalized into the det.-rel. pronoun, while the adnominal marker *ha([+ or -]n) was grammaticalized into the definite article in Hebrew, Aramaic, and OSA. On the basis of this typological interpretation it can be hypothesized that the demonstrative pronouns in PS had the same base in the singular as in PWS and were case-inflected just as the det.-rel. pronouns. Case was lost in PWS, since it is the lowest category in the inflectional chain pertaining to demonstratives. The pl. base can be reconstructed as * ul for PS, based on its use for the plural in WS and its existence in Akkadian, where it functions as a remote demonstrative in Babylonian.
The investigation of demonstratives in Semitic is thus a good example of a situation where language typology can be helpful for language reconstruction.
Allegro, J. M. 1955. Uses of the Semitic Demonstrative Element Z in Hebrew. VT 5: 309-12.
Allen, J. P. 2000. Middle Egyptian: An Introduction to the Language and Culture of Hieroglyphs. Cambridge: Cambridge Univ. Press.
Barth, J. 1907. Sprachwissenschaftliche Untersuchungen zum Semitischen. Erster Teil. Leipzig: Hinrichs.
--. 1913. Die Pronominalbildung in den semitischen Sprachen. Leipzig: Hinrichs.
Bauer, H., and P. Leander. 1922. Historische Grammatik der hebraischen Sprache. Rpt. Hildesheim: Georg Olms, 1991.
Beeston, A. F. L. 1984. Sabaic Grammar. Manchester: Univ. of Manchester.
Behnstedt, P. 1993. Die demonstrativen Bildungen der syrisch-arabischen Dialekte. Zeitschrift fur arabische Linguistik 25: 76-94.
Brockelmann, C. 1908. Grundriss der vergleichenden Grammatik der semitischen Sprachen, vol. 1. Berlin: Von Reuther & Reichard.
Brustad, K.F. 2000. The Syntax of Spoken Arabic: A Comparative Study of Moroccan, Egyptian Syrian, and Kuwaiti Dialects. Washington, D.C.: Georgetown Univ. Press.
Chiera, G. 1996. II pronome zu in ebraico. In Alle soglie della classicita: Il Mediterraneo tra tradizione e innovazione: Studi in onore di Sabatino Moscati, vol. 3, ed. E. Acquaro. Rome: Instituti Editoriali Polografici Internazionali. Pp. 1109-15.
Christian, V. 1924. Die deiktischen Elemente in den semitischen Sprachen nach Herkunft, Anwendung und Verwandtschaft untersucht. WZKM 31: 137-92.
Dalman, G. 1927. Grammatik des Judisch-Palastinischen Aramdisch. Rpt. Darmstadt: Wissenschaft-liche Buchgesellschaft, 1981.
Degen, R. 1969. Altaramaische Grammatik. Wiesbaden: Franz Steiner Verlag.
Diakonoff, I. 1988. Afrasian Languages. Moskow: Nauka.
Diessel, H. 1999. Demonstratives: Form, Function, and Grammaticalization. Amsterdam: John Benjamins.
Ehlich, K. 1979. Verwendung der Deixis beim sprachlichen Handeln: Linguistisch-philologische Untersuchungen zum hebraischen deiktischen System, Teil 2. Frankfurt: Peter Lang.
Fischer, W. 1959. Die demonstrativen Bildungen der Neuarabischen Dialekte. The Hague: Mouton.
--. 1987. Grammatik des klassischen Arabisch. Wiesbaden: Harrassowitz.
Fischer, W., and O. Jastrow. 1980. Handbuch der arabischen Dialekte. Wiesbaden: Harrassowitz.
Friedrich, J.; W. Rollig; and M. G. Amadasi Guzzo. 1999. Phonizisch-Punische Grammatik. 3rd ed. Rome: Pontifical Biblical Institute.
Fronzaroli, P. 1987. Le pronom determinatif-relatif a Ebla. MARI 5: 267-74.
Gesenius, W.; E. Kautzsch; and A. E. Cowley. 1910. Gesenius' Hebrew Grammar. Oxford: Clarendon Press.
Haile, G. 1967. Demonstrative Pronouns in Amharic. Journal of Ethiopian Studies 5: 9-12.
Hasselbach, R. 2005. Sargonic Akkadian: A Historical and Comparative Study of the Syllabic Texts Wiesbaden: Harrassowitz.
Heine, B., and M. Reh. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske.
Huehnergard, J. 1998. A Grammar of Akkadian. Atlanta: Scholars Press.
--. 2006. On the Etymology of the Hebrew Relative s[epsilon]. In Biblical Hebrew in its Northwest Semitic Setting: Typological and Historical Perspectives, ed. S. E. Fassberg. Winona Lake, Ind.: Eisenbrauns. Pp. 103-25.
Johnstone, T. M. 1975. The Modern South Arabian Languages. Afroasiatic Linguistics 1: 1-29.
Kienast, B. 2001. Historische semitische Sprachwissenschaft. Wiesbaden: Harrassowitz.
Kogan, L. 1997. Tigrinya. In The Semitic Languages, ed. R. Hetzron. New York: Routledge. Pp. 424-45.
Kurytowicz, J. 1945-49. La nature des proces dits "analogiques." Acta Linguistica 5: 15-37.
Lambdin, T. O. 1978. Introduction to Classical Ethiopic (Ge'ez). Atlanta: Scholars Press.
Lipinski, E. 2001. Semitic Languages: Outline of a Comparative Grammar. Second edition. Louvain: Peters.
Lonnet, A. 2003. L'accumulation des deictiques: L'expression de 'maintenant' en sudarabique moderne. In Melanges David Cohen: Etudes sur le langage, les langues, les dialectes, les litteratures, offertes par ses eleves, ses collegues, ses amis, ed. J. Lentin and A. Lonnet. Paris: Maisonneuve et Larose. Pp. 421-38.
Macdonald, M.C.A. 2004. Ancient North Arabian. In The Cambridge Encyclopedia of the World's Ancient Languages, ed. R. D. Woodard. Cambridge: Cambridge Univ. Press. Pp. 488-533.
Moscati, S., et al. 1964. An Introduction to the Comparative Grammar of the Semitic Languages. Wiesbaden: Harrassowitz.
Muraoka, T., and B. Porten. 2003. A Grammar of Egyptian Aramaic. 2nd rev. ed. Leiden: Brill.
Nebes, N., and P. Stein. 2004. Ancient South Arabian. In The Cambridge Encyclopedia of the World's Ancient Languages, ed. R. D. Woodard. Cambridge: Cambridge Univ. Press. Pp. 454-87.
Noldeke, T. 1904. Compendious Syriac Grammar. Rpt. Winona Lake, Ind.: Eisenbrauns, 2001.
O'Leary, de Lacy. 1923. Comparative Grammar of the Semitic Languages. Rpt. Amsterdam: Philo, 1969.
Pardee, D. 2004. Ugaritic. In The Cambridge Encyclopedia of the World's Ancient Languages, ed. R. D. Woodard. Cambridge: Cambridge Univ. Press. Pp. 288-318.
Pennachietti, F. A. 1968. Studi sui pronomi determinativi semitici. Napoli: Istituto Orientale di Napoli.
--. 2005. Ripercussioni sintattiche in conseguenza dell' introduzione dell' articolo determinativo proclitico in semitico. Aula Orientalia 23: 175-84.
Rubin, A. 2005. Studies in Semitic Grammaticalization. Winona Lake, Ind.: Eisenbrauns.
Raz, S. 1983. Tigre Grammar and Texts. Malibu: Undena.
--. 1997. Tigre. In The Semitic Languages, ed. R. Hetzron. New York: Routledge. Pp. 446-56.
Saenz-Badillos, A. 1996. A History of the Hebrew Language. Cambridge: Cambridge Univ. Press.
Simeone-Senelle, M.-C. 1997. The Modern South Arabian Languages. In The Semitic Languages, ed. R. Hetzron. New York: Routledge. Pp. 378-423.
von Soden, W. 1995. Grundriss der akkadischen Grammatik. 3rd ed. Rome: Pontifical Biblical Institute.
Testen, D. 2005. The Akkadian Demonstrative ammiu. In Studia Semitica et Semitohamitica: Festschrift fur Rainer Voigt anlafilich seines 60. Geburtstages, ed. B. Burtea et al. Munster: Ugarit-Verlag. Pp. 405-16.
Tropper, J. 2000. Ugaritische Grammatik, Munster: Ugarit-Verlag.
Voigt, R. M. 1988. Die Personalpronomina der 3. Personen im Semitischen. WO 18: 49-63.
Wright, W. 1898. A Grammar of the Arabic Language. Rpt. Cambridge: Cambridge Univ. Press, 1967.
Zaborski, A. 1984-86. A Note on Cushitic Demonstrative Pronouns. Orientalia Suecana 33-35: 505-11.
I would like to thank sincerely N. Pat-El and A. Rubin for their helpful comments and suggestions for improving this essay. All opinions and errors are, of course, mine alone. Abbreviations used in this article are: AA = Afro-Asiastic; acc. = accusative; ANA = Ancient North Arabian; anaph. - anaphoric; BA = Biblical Aramaic; c = comunis; CA = Classical Arabic; CPA = Christian Palestinian Aramaic; CS = Central Semitic; d = dual; det.-rel. = determinative-relative; ES = East Semitic; f = feminine; fem. = feminine; gen. = genitive; JPA - Jewish Palestinian Aramaic; m = masculine; masc. = masculine; MSA = Modern South Arabian; nom. = nominative; OB = Old Babylonian; obl. = oblique; OSA = Old South Arabian; p = plural; PS = Proto Semitic; s = singular; WS = West Semitic.
(1.) For a more complete list, see section 3 below.
(2.) The transcription with [delta] represents the PS phoneme *[delta], while d indicates the secondary spirantization of /d/ after vowels in Hebrew and Aramaic.
(3.) Kienast (2001:50).
(4.) See Brockelmann 1908: 107a; O'Leary 1923: 58; Christian 1924: 137. This derivation necessarily presumes that the various demonstrative particles have no original connection to gender or number (ibid.). For the well-known preservation of ha as an interjection in Semitic languages such as Arabic and Aramaic, see Brockelmann (ibid.); Barth 1913: 72-74.
(5.) Barth 1913: 72.
(6.) Ibid., 74.
(7.) Barth 1907: 19-20.
(8.) Fischer and Jastrow 1980: 82.
(9.) Barth 1913: 77-78. For the connection of Arabic 'alla[delta]i and Hebrew hallaze, see also Bauer and Leander 1922: [section]30g.
(10.) Barth 1913: 96. The final long -a in the biform -na in Aramaic has been interpreted as the definite article (Brockelmann 1908: [section]107v). This analysis is not convincing, since final /a/ is also attested in languages that do not have a definite article of this type, such as Maltese (for the attestations in Maltese see section 3.7 below).
(11.) Barth 1907: 31; 1913: 83.
(12.) Brockelmann 1908: [section] 107t; similarly also Kienast 2001: 49.
(13.) East Cushitic uses two main demonstrative bases, k- and t-, although also in these languages t- and k- primarily distinguish gender, not spatial concepts (Zaborski 1984-86: 505).
(14.) Oblique independent pronouns with /t/ are known from Babylonian Akkadian, Eblaite, OSA, and Ugaritic. Vestiges of an oblique t-pronoun and found in Phoenician. The distribution of t = oblique vs. [empty set] = nom. is probably original. Some languages subsequently leveled the t-containing pronouns, as in Assyrian and Ge'ez, while others, such as Arabic, Hebrew, and Aramaic, leveled the originally unmarked nom. forms. A similar interpretation is found in Lipiriski 2001: 308.
(15.) Brockelmann 1908: [section]107e and o. The expansion particle in MSA, interpreted by Lonnet as deictic suffix (2003: 434), is used for other deictic particles as well, such as 'here,' Mehri both and bumah, 'there,' Mehri hok and helakemeh (Lonnet 2003: 430).
(16.) O'Leary 1923: 160, 168; Diakonoff 1988: 83.
(17.) Barth 1913: 103.
(18.) The same change presumably underlies Syriac hade (fs) (Barth 1907: 31, 36; 1913: 86). The secondary character of the Arabic distribution is quite likely and can be confirmed by the ms relative pronoun 'alla[delta]i, which has an i/ instead of the expected /a/ vowel--gender distinction being indicated by the consonant and not the vowel in the singular, that is 'alla[delta]i (ms) ~ 'allati (fs). The distinction based on a consonantal change might have prevented the secondary distinction by vowel Ablaut. For the relationship of the relative pronoun and the demonstratives see also section 4 below.
(19.) Fischer 1959: 35 n. 1.
(20.) Brockelmann 1908: [section]107t. The same analysis is found in Bauer and Leander 1922: [section]30e.
(21.) Huehnergard 2006: 114.
(22.) Barth 1913: 118; Bauer and Leander 1922: [section]30d.
(23. Lipiriski 2001: 323-24.
(24.) Diessel 1999.
(25.) Ibid., 4-5.
(26.) Ibid., 25.
(27.) Ibid., 33.
(28.) Ibid., 119-28; Heine and Reh 1984: 271.
(29.) The etymology of Assyrian ammium is still a matter of dispute. In a recent article Testen suggested that ammium might be derived from a PS /h/ - containing 3mp independent pronoun * hmu > * hamu > * amu, with the consonantal gemination and final -i resulting from an analogy with common Semitic ulli (Testen 2005: 409, 413). This derivation is extremely unlikely since it has been argued convincingly that third person independent pronouns did not have an /h/-containing form in PS; for a comprehensive treatment of this topic, see Voigt 1988. It would perhaps be more fruitful to compare ammium with the expansion particle -(e)meh used on demonstrative and deictic pronouns in MSA languages, although this requires further study.
(30.) Von Soden 1995 (3): [section]45a; Huehnergard 1998: 41.
(31.) Lambdin 1978: 30.
(32.) For the attestations see Raz 1983: 45; 1997: 451.
(33.) Kogan 1997: 434.
(34.) Johnstone 1975: 24; Simeone-Senelle 1997: 394; Lonnet 2003: 428-29.
(35.) Simeone-Senelle 1997: 413.
(36.) Beeston 1984: 41; Nebes and Stein 2004: 463.
(37.) Beeston 1984: 32. According to Beeston, Sabaic has vestiges of case inflection, such as the distinction of nom. ~ non-nom. forms in the far demonstratives, although it seems that case distinction was lost during the time in which Sabaic is attested.
(38.) Ibid., 40-41.
(39.) Macdonald 2004: 508-9.
(40.) Ibid., 508.
(41.) According to Wright, some Arab grammarians noted a semantic difference between the remote demonstratives with and without -li. The form without infix supposedly refers to the nearer object, while the form with -li refers to the more remote one when two objects are compared (Wright 1898: 267).
(42.) A rare biform that was nearly extinct already in classical times is the older fem. ha[delta]i (Fischer 1987: [section]274).
(43.) The doubling of /n/ in these forms presumably derives from *[delta]anlika, with an irregular regressive assimilation of /l/ to /n/ (Fischer 1987: [section]275). Rarely, there also occur forms of far demonstratives with prefixed ha-, such as ha[delta]aka (ibid.).
(44.) Fischer 1959: 35, 59, 86; 1987: [section]274. The forms with ha-, seem to have originally been more "emphatic" than the forms without ha-. Consequently, ha-. is called "the particle that excites attention" by Arab grammarians (Wright 1898: 268; Fischer 1959: 37). In some modern Arabic dialects, ha- still has real demonstrative force, although it was lost in others such as Palestinian, Syrian, and Lebanese Arabic (Fischer 1959: 51).
(45.) Exceptions to this general rule are the genitive construction where the demonstrative has to follow when it modifies the first noun, and a noun plus pronominal suffix, as in baytu rrajuli ha[delta]a 'this house of the man' (as opposed to baytu ha[delta]a rrajuli 'the house of this man'), and baytuhu ha[delta]a 'this house of his.'
(46.) Fischer 1959: 41; Behnstedt 1993: 76. Fischer does not consider this ha- in modern Arabic dialects as being derived by shortening of the Classical Arabic form ha[delta]a, but as being independently derived from the interjection ha (1959: 47). According to Fischer, compounds such as halli < *ha-illi originally meant 'there is the one who ...,' with subsequent semantic bleaching leading to the interpretation as a simple near demonstrative (1959: 51). Thus, according to Fisher, the forms with ha-originally expressed a more intensive notion than the simple demonstratives: "Eine Verschiebung der ursprunglichen Funktion ist nur insofern eingetreten, als heute die ha- Formen die normalen Demonstrativa darstellen und die Formen ohne ha ausgesprochene Schwachformen sind, wahrend zunachst wohl da, dak als Normalform des Damonstrativpronomens fungierte und ihnen das prafigierte ha eine besonders affektische Bedeutung verlieh" (Fischer 1959: 37).
Brustad interprets the use of gender- and numberless prefixed ha- in modern Arabic dialects as functionally different from the inflected and often post-nominal demonstrative pronouns. The prefixed form has anaphoric function, that is, it refers back to a previously mentioned entity, while the independent demonstrative pronoun has a deictic function (Brustad 2000: 113). In Moroccan Arabic, we also find "double" demonstrative constructions with prefixed had--equivalent to ha- in other dialects--and a demonstrative pronoun in post-nominal position. According to Brustad, this construction expresses both anaphoric presence and a "heightened specificity," as in had l-qadiyyo. hadi 'this particular problem' (2000: 132).
(47.) Fischer 1959: 50; Fischer and Jastrow 1980: 82.
(48.) Fischer 1959: 67. Fischer considers the final -n in Maltese a secondary development, not the preservation of a feature that would be related to demonstrative forms in OSA and Aramaic (1959: 69).
(49.) Fischer and Jastrow 1980: 82. Most of the modern Arabic dialects have the base *'ul for the plural, although there are a few dialects which seem to have a base with an /i/- vowel *'ila, reflected in forms such as dayla and del(a) (ibid.).
(50.) Fischer 1959: 207; Brustad 2000: 114. These are some of the forms attested in Moroccan diales. [delta]a(k) / 5l(k) arc often used for both genders, although some areas distinguish between [delta]a for masc. and [delta]l for fern.; the forms with prefixed ha- are generally distinguished for gender (ibid.: 57). The following attestations are taken from Fischer 1959:207-21.
(51.) Fischer suggests that forms with final /'/or/h/ were pausal forms that were extended to non-pausal contexts (Fischer 1959: 59). The final /i/ on ha[delta]ihi presumably is a svarabhakti-vowel that originally appeared when the demonstrative was followed by the definite article (ibid.).
(52.) Ibid., 53, 57.
(53.) Exceptions to this syntactic flexibility are certain constructions such as genitive constructions, in which the demonstrative has to follow the noun (Brustad 2000: 113).
(54.) Pardee 2004: 301.
(55.) Tropper 2000: 229-30.
(56.) Ibid., 231-32.
(57.) The form zo(')t is the standard fs demonstrative in Biblical Hebrew, while the other two are rare biforms.
(58.) The biform 'el is only used in the Pentateuch and only eight times. It always appears in combination with the definite article as ha'el (Gesenius, Kautzsch, Crowley 1910: [section]34b).
(59.) Bauer and Leander consider zu a demonstrative used only in poetry (Bauer and Leander 1922: [section]30e). Ehlich, in his study of deixis in Hebrew, comes to the conclusion that zu is exclusively used as relative pronoun (Ehlich 1979: 639)--for the interpretation of zu as expressing a genitival relationship see also Allegro (1955: 311)--while Chiera states that zu sometimes corresponds to a demonstrative (1996: 1114).
(60.) Saenz-Badillos 1996: 186.
(61.) Friedrich, Rollig, Amadasi-Guzzo 1999: [section]113.
(62.) Ibid., [section]217.
(63.) The forms z and zn are found in Byblos, and z' in Cyprus, although it is only attested once. The prosthetic syllable 'z was probably vocalized as /'e/ (ibid., [section]116).
(64.) The form z' is only attested once for the fem. in a text from Byblos (ibid., [section]115). The forms z and 'z are used for both the masc. and fem.
(65.) Ibid., [section]113.
(66.) Ibid., [section]288.
(67.) See Degen 1969: 59.
(68.) z't is attested in Fekheriye, 'l in the Zakkur inscription, 'ln in Sefire.
(69.) Muraoka and Porten 2003: 56-57.
(70.) dk' occurs only once and is probably a scribal error for dky (ibid., 57 n. 275). The forms with -ky at the end often stand for the fem., while forms without -y are commonly used for the masc, although this distribution is not consistent (166).
(71.) Ibid., 166.
(72.) Dalman 1927: 396-97. Similar forms are found in CPA and in Targum Onqelos. Forms in parentheses are rare.
(73.) This form is derived from hainnun, a form attested in the Palestinian Targums.
(74.) Dalman 1927: 113-14.
(75.) Noldeke 1904: [section]226.
(76.) This statement is based on the assumption that ze- and zentu in Ge'ez are derived from *zi- and *zintu respectively. It is theoretically possible to assume a derivation from **zu- and **zuntu since both short/i/and short/u/merged to/e/in Ge'ez, but this is unlikely since it is much more common to have an a~i Ablaut between the masc. and fem. in WS than an a~u Ablaut. The vowel quality in Aramaic is based on forms like den, which have to be derived from *din. In Hebrew, an original long vowel should have yielded **zi, not ze.
(77.) Fischer 1959: 61.
(78.) Rubin 2005: 76. For han < ha+n see Moscati 1964: 112.
(79.) See also section 1 above.
(80.) Late Aramaic dialects show a more complex situation. JPA has a remote demonstrative hahu('), which is probably a loan from Hebrew. Furthermore, there are forms that go back to a symbiosis of the particle ha- and the anaphoric pronouns, as, for example, Syriac haw < *ha+hu and hannon < *ha+'ennon.
(81.) Kurylowicz 1945-49.
(82.) Such a comparison would explain why Hebrew uses the definite article when the demonstrative is used adnominally, since ha([+ or -]n) originally marked adnominal function (see section 4 below for this interpretation of ha([+ or -]n)).
(83.) Pennachietti explains the use of the definite article on attributive adjectives in Semitic languages that make use of a proelitic definite article as similar to the use of a det.-rel. pronoun, that is, hak-kohen hag-gadol would mean something like "the priest, the one who is big." The definite article on the attributive adjective would thus have an anaphoric function (Pennachieetti 2005: 178). According to Pennachietti, the definite articles derived from the demonstrative base *han- / * hal-, and the original det.-rel. pronoun that was equally derived from a demonstrative base, *[delta]V:, became incompatible in those languages that developed a proclitic definite article, partly because of an overlap in function, which ultimately led to the loss of the original det.-rel. pronoun *[delta]V: in these languages (2005: 179, 183).
(84.) See Diessel 1999: 33, who states that inflectional agreement between pronominal demonstratives and nouns is a general typological characteristic: "In languages in which nouns are inflected for gender, number and/or case, pronominal demonstratives are always marked for the same features whereas adnominal and identificational demonstratives are often uninflected."
(85.) For the relationship of [delta]V: and [theta]: despite their diverging initial consonants, see Huehnergard 2006: 114. It is not clear whether Eblaite had a voiceless initial consonant like Akkadian or the voiced variant found in WS, since the Eblaite writing system uses the same sign series for both *[delta] and *[theta] (ibid.).
(86.) The vowel length in the singular is not certain. It is given as short based on later Akkadian sa. Eblaite forms that differ from the Old Akkadian ones or forms that are only attested in Eblaite are indicated in parentheses. For Old Akkadian, see Hasselbach 2005: 161; for Eblaite Fronzaroli 1987: 267-73.
(87.) Fischer 1987: 131.
(88.) The uniform ending -a in rel. pronouns in Ge'ez reflects the construct ending.
(89.) In Hebrew, these are only vestigial forms primarily used in poetic texts. See also Huehnergard 2006: 111.
The common relative pronoun used in Biblical Hebrew is "(a)ser, which is not inflected for gender and number. For a general discussion of these forms, sec also Pennachietti 1968: 8-11.
(90.) Huehnergard 2006: 112.
(91.) Ibid., 112-13 n. 59.
(92.) Ibid., 112-13 n. 59.
(93.) For a recent discussion and literature, see Rubin 2005: 65-79.
(94.) Rubin 2005: 72-80. In my opinion, Rubin's reconstruction is quite convincing on a morphological level and will be followed in this study.
(95.) It is unlikely that the adnominal function of ha-developed out of the use of the definite articles, first, because it would contradict general typology, and, more importantly, because Akkadian seems to have generalized the adnominal form without developing a definite article.
(96.) For Egyptian, see Allen 2000: [section]:5: 8-9. In Middle Egyptian, the suffix -n is combined with the three bases for masc, fem. and neuter: p-, t-, and n- respectively, resulting in the forms pn, tn and nn. All these demonstratives are used for both near and far deixis and can be used both substantivally and attributively. The other Egyptian demonstrative suffixes, -w, f, . and -' have no equivalent in Semitic. Cushitic languages likewise have a demonstrative element n used for non-distant objects, as do Berber languages such as Kabyle (Zaborski 1984-86: 505).
(97.) Whether the suggested adnominal function of -n can be connected to the nominal marker -n commonly known as nunation is not certain, although an association with this morpheme might explain how an adnominal marking function of -n developed within Semitic.
(98.) Haile 1967: 9-12.
(99.) Zaborski 1984-86: 508.
(100.) A nominal pattern *quttil is not attested in any Semitic language, although it is known in verbal formations in Akkadian and Arabic.
(101.) The derivation of relative pronouns and markers from demonstratives is a very well-known phenomenon cross-linguistically, while the opposite development is unattested. See, e.g., Heine and Reh 1984: 271, 281.
(102.) From a typological perspective, grammaticalization results in the reduction of members belonging to a given paradigm. Thus it would be hazardous to assume that the case distinction of the Semitic det.-rel. pronoun is a secondary development that occurred after the grammaticalization from the demonstrative pronouns. For this analysis see Heine and Reh 1984: 67.
(103.) The assumption that PS had a plural base * ul does not contradict the fact that the ES relative pronoun does not reflect this base. Grammaticalization does, as mentioned above, result in the reduction of paradigmatic variation. The loss of* ul in the relative pronoun is an example of this kind of reduction.
UNIVERSITY OF CHICAGO