The historical linguistics of the intrusive *-n in Arabic and West Semitic.
I term the -n (realized variously as -in, -an, -inn, -ann, -unn, -anna) that occurs before object suffixes in many Semitic languages and varieties the "intrusive -n," or simply "-n." (1) The origin of this morpheme has been widely discussed among Semiticists in particular, and more recently among Arabicists (Holes 2011). Among Semiticists there are two broad explanations for its appearance. The more widespread approach, represented inter alia by Robert Hetzron (1969), David Testen (1993), and Rebecca Hasselbach (2006), is to interpret the intrusive -n as an inherited proto-Semitic verbal suffix with various, uninterrupted reflexes across the different varieties. An alternative perspective is offered by Jan Retso (1988: 92; also Barth 1907), who sees the -n as originating independently from common "deictic elements," a massive parallel independent development, as it were. The former approach tends to derive the morpheme from common verbal aspect-mode values, two functions of which are the Akkadian ventive (motion towards speaker) and the Classical Arabic so-called energicus (nun al-ta'kid in the Arabic tradition, also termed energetic or energic in the Western tradition). The latter approach, on the other hand, is sceptical of shared proto-functions, instead emphasizing the basically formal property of pronoun object marking.
In this article I will attempt to combine elements of each of the two perspectives. As in the first position, there is a common, shared origin to all occurrences of the -Vn in West Semitic. However, following Retso, the only functional unity that characterizes them is the grammaticalized function of marking a pronoun object suffix. As will be seen, the explanation advocated here basically confirms the analysis of Carlo Landberg (1909: 738). (2)
Partly for strategic reasons, and partly because of the breadth of material that would have to be treated in detail to work out a completely comprehensive development, I will concentrate on the West Semitic languages in general, and on Arabic in particular. To the extent that the West Semitic languages have the intrusive -n, they are remarkably similar to one another in grammar, while Arabic is of particular interest for two reasons. First, the -n is attested in a number of varieties of it, and at different chronological eras; and second, a close look at the general syntax of the -n in a major text of Classical Arabic will allow a detailed evaluation, and refutation, of the idea that the -n derives from a proto-verbal function.
I begin in section 2 with a summary of the situation in Arabic, first the dialects, then a corpus-based summary of the energic -n in Qur'anic Arabic. This will form the basis in sections 2.2 and 2.3 of the first of two reconstructions developed in this paper, namely, a reconstruction of the development of the energic as a grammatical category in Classical Arabic. In section 3 the data from other West Semitic languages are presented, and in section 4 an overall historical development is offered. In section 5 individual interpretive issues related to the proposed solution are addressed.
2. THE INTRUSIVE -N IN ARABIC
2.1. Arabic dialects
The properties of -n are quite uniform among contemporary Arabic dialects. In all Arabic dialects where it occurs--viz., Eastern Arabian dialects of former South Yemen, Oman, the Emirates, and Bahrain; those of eastern Syria, Khorasan, Uzbekistan, and Afghanistan; and Bagirmi Arabic in northeast Nigeria, Cameroon, and western Chad--an intrusive -n is added onto the active participle only before an object suffix.
In Yemen (specifically, Dathina, in former South Yemen; cf. Landberg 1909: 720ff.) it occurs in all forms of the AP. The -n is geminate except before the -n initial suffixes -ni 'me' and -na 'us'. Gemination nearly always occurs before a vowel, although there are a few examples of -nn-ha (mayaahib-Inne-ha "he has accompanied her," as well as mithaddin-inn-ha "he carried her in the arms"; subjects are provided from context in the following examples):
(1) mehaalif-inn-ak "[he is] your ally" (allied-N-you-M)
muhaalif-iin-in-na "[they are] allied to us" (allied-MPL-N-us)
kaatib-et-inn-eh "[you (F)] have written it (M)"
kaatb-aat-inn-a "[they (F)] have written it (M)"
In Oman -in is added to any stem form (Reinhardt 1972: 139): (3)
(2) daarb-Inn-ek "he has hit you (M)"
daarb-ft-n-hum "she has hit them (M)"
daarb-fin-n-ek "they (M) have hit you (M)"
daarbaat-inn-is "they (F) have hit you (F)"
Here -n is added to the MPL directly, rather than as -in + pronoun suffix, as in Yemen.
In Bahrain (cf. Holes 1987: 109), -in is a feature of Shi'ite (Bahama) speakers. It is added only after a singular participle (M or F); -n is geminate before a vowel, single before a consonant: (4)
(3) xaatb-in-ha "he has become engaged to her"
xaatb-at-inn-ah "she has become engaged to him"
In the Syrian desert I. G. Wetzstein notes (1868: 192; no complete paradigms are given in this work) that forms with and without -n are used before object suffixes:
(4) saayf-ann-u "he saw him" (1868: 75)
For Khorasan Ulrich Seeger (2002: 635) does not give complete paradigms, but he shows the -n suffixed to the AP stem for MSG, and to the FSG suffix for the feminine singular. He gives only a geminate -nn, except for n- initial suffixes (as with Oman and Yemen), and he has u for the vowel.
(5) mint-unn-he "he has given her" (give-N-her)
aaxd-t-unn-a "I (F) took [married] him" (take-F-N-him)
In Uzbekistan and Afghanistan (5) the situation is slightly complicated by the fact that the active participle has itself become refunctionalized into a person-inflected form in which the historical pronoun objects assume the function of subject. However, this refunctionalized pronoun object is always suffixed to the -n in the first and second person subjects, (6) so that formally, even if not semantically, it parallels the other dialects. The -n suffixes directly to the MSG AP stem, neutralizing FSG and MPL suffixes. For Afghanistani Arabic Bruce Ingham (2006: 32) gives the following forms, some, it can be noted, with -an instead of -in, while the 1 and 2MPL forms are suffixed to the plural suffix -iin, rather than the intrusive -n.
(6) SG PL 1 kaatb-an-ni (7) kaatb-iin-na 2M kaatb-inn-ak kaatb-iin-ku 2F kaatb-an-ki kaatb-an-kin
The -n is geminate before -V, single otherwise, a situation also noted by Wolfdietrich Fischer (1961) for Uzbekistan.
For Uzbekistan, Gent Zimmermann's (2009: 620) data agree broadly with (6), except that she clearly has the -n in all forms including the plurals--qard-in-kum "you (MPL) have sat down"--and always a single -n.
Finally, in Bagirmi Arabic, a dialect area roughly coterminous with the former Bagirmi empire that ruled the border area between western Chad, Cameroon, and Nigeria south of
Lake Chad, -n is added to the MSG stem, where, as with Uzbekistani Arabic, it neutralizes all other forms (Owens 1993: 2009).
(7) ana kaatb-in-ha "I have written it" vs. ana kaatib (M) "I have written"
inti kaatb-in-ha; inti kaatb-e "you (FSG) have written"
hum kaatb-in-ha; hum kaatb-iin "they (M) have written"
hinna kaatb-in-ha; hinna kaatb-aat "they (F) have written"
In addition, in all non-Bagirmi Nigerian Arabic dialects and in Shukriyya, in eastern Sudan (Reichmuth 1983), -an/-ann is added to the FPL -aat, as in Oman (see (2), above). The geminate variant occurs before a vowel, the single before a consonant.
(8) kaatb-aat-ann-u "they (F) have written it" (written-FPL-N-itM)
In these two dialects, the -an can probably be interpreted as a relic of the fuller system, as in (1) and (2), above.
2.1.1. Other -n/-nn's
In the eastern Arabian peninsula, southeastern United Arab Emirates, and the adjoining area of Oman, an -n can be suffixed between pronoun object and imperfect verb (Holes 2011). Clive Holes notes that such forms are common in this area: (8)
(9) yi-kaffi-n-na "it will be enough for us" (3M-enough-N-us)
n-sawwi-nn-a "we do it" (we-do-N-itM)
yi-Sill-inn-ah "he removes it" (3M-remove-N-itM)
The cognate status of these is discussed in section 5.4, below.
It is relevant, for reasons that emerge in sections 2.3 and 4, below, to briefly summarize another -n, this one occurring in a class of complementizers in Arabic, which in Classical Arabic are those ending in -nna (inna wa-akhawatuha), namely, inna 'that', anna 'that', and lakinna 'but'. These have a number of formal links to the energic, discussed in the next two sections. Like the energic they can be followed by a pronoun from the object series, (9) anna-ka 'that you'; they are neutrally followed by an accusative complement if nominal rather than pronominal; and anna itself marks an object complement.
(10) 'alim-tu anna zayd-an muntaliq-un learn-I that Zayd-ACC leave-NOM "I learned that Zayd is leaving"
Furthermore, there is a "light" version of this morpheme (in, an, lakin) that in most varieties does not govern an accusative noun (Sibawayhi 1970, 1: 430).
The complementizer in continues to be used in many contemporary dialects, though not all (see 5.6, below). It has, however, various functions, only one of which is to mark a sentential verb complement (e.g., Germanos 2009, where it is a discourse marker, as well as marking complements). Retso (1988: 79) notes that a number of Landberg's texts have a particle win, often used with a suffix pronoun co-referential with the subject of the clause, which appears to demarcate short episode segments in narratives.
(11) winne-hom nezal-u min em-xalwah leoneen Cala shaab-hom ... u qaal-u inna tacan-na Oneen u elcalim alla inne-hom faat-u oo saad-hom
"and then the two descended to their companions ... and they said we stabbed two and God knows if they are dead or not" (Landberg 1905: 9) (10)
As can be seen here, inna occurs in the same texts marking a sentential complement after qaal, so that there appear to be multiple *-in-origin words in this dialect.
2.2. Classical Arabic
Following a long tradition of scholarship, it may be assumed that the -n of the Arabic AP is of the same historical lineage as the energic -n of Classical Arabic. As will be shown in section 2.3 it can be more precisely argued that the energic is an innovative category built on an older -n construction.
In Classical Arabic there are two morphemes that in the Western tradition are known as the energic, -anna and -an. These are termed in the Arabic tradition al-nan al-thaqila and al-nun al-khafifa respectively, or the heavy and the light -n. They are suffixed to the imperfect and imperative verbs. According to the explanation of the grammarian Sibawayhi (d. ca. 180/796), who wrote the earliest Arabic grammar, in the 2FSG and 2/3MPL forms the long vowels of the suffixes shorten regularly according to the general rule that shortens long vowels before -CC, while the initial -a of the energic is elided. (11) The FPL produces the allomorph -anni:
(12) SG PL a-ktub-anna na-ktub-anna ta-ktub-anna ta-ktub-u-nna ta-ktub-i-nna ta-ktub-n-anni ya-ktub-anna yaktub-u-nna ta-aktub-anna ya-ktub-n-anni
Further special variants occur with the dual, which are not relevant to this discussion.
The distribution of the "light" -n (termed energic II in the Western tradition) is by comparison curtailed. The issues are complex and only basic facts will be summarized here. Sibawayhi spends nine pages in his grammar al-Kitab describing the two -n's (1970, 2: 152-61), with much of the description devoted to the various allomorphic or variationist realizations of the light -n, including detailed discussion and criticism of the views of Yunus Ibn Habib (d. 182/798) (and the nahwiyyun) on the subject. The light -n does not occur in the dual or the feminine plural. Furthermore, it does not occur before either the definite article 1- or an epenthetic vowel (2: 157.6, 158.17), since this would lead to an unacceptable sequence of three consonants (*tukrim-an l-dayf-a "you honor the guest"). (12) In pausal position it changes to -a after an a. In the 2FSG and plural forms Sibawayhi notes that the -n of the indicative (ran returns in pausal position, i.e., instead of theoretical taktub-i-n "you (F) write-N" > taktub-iy-a or taktub-in-a, one has taktub-in (2: 158.9).
(13) SG PL aktub-an na-ktub-an taktub-an ta-ktub-u-n (but not in pause) taktub-i-n 0 yaktub-an ya-ktub-u-n (but not in pause) taktub-an 0
The energic is a translation of the term tawkid 'emphasis', a term that Sibawayhi attributes to al-Khalil (d. 175/791) (2: 152). Sibawayhi notes that the heavy -n is more emphatic than the light, and he gives two contexts with which it is closely associated. The first is the so-called lam al-qasam--the lam or letter I signifying an oath--a morpheme that he sees as closely associated with the energic (see Testen 1998: chaps. 1 and 2 for summary of la- in Arabic). Sibawayhi explains qasam 'oath' as "emphasizing your expression" (ta'kid li-kalamika, 1: 403.15), i.e., it is a linguistic, not a legal category. Should la- be used with a non-negative imperfect verb, Sibawayhi claims that the energic must also be used:
(14) wa-llah-i la-a-f'al-anna by God-GEN EM-I-do-T "By God! will do [it]" (2: 403.18)
This formulation is too general, as in fact la- does occur without the energic, as Sibawayhi himself later notes:
(15) inna zayd-an la-ya-drib-u EM Zayd-ACC EM-3-hit-IND "Indeed Zayd will hit" (1: 405.14)
Sibawayhi simply observes that the use of the energic with la- here is more frequent. What can be noted is that Sibawayhi often illustrates the energic in the context of a lexical expression with the force of an oath, wa-llah 'by God', ashhadu 'I bear witness', aqsimu `I swear', and the like (see (29), below).
A second context is in imperative and negative verbs, la ta-f'al-anna 'don't do!', which Sibawayhi notes as optionally occurring with the energic (in shi'ta adkhalta fihi 1-nun wa-in shi'ta lam tudkhil). In addition Sibawayhi summarizes further contexts, all characterized by occurrence with a characteristic but optional morpheme. One of particular interest is his observation that with a question word the energic turns a question into a command, i.e., it assumes the illocutionary force of a command:
"do you really mean to say?" = "you don't (really) mean to say; you shouldn't say"
In this context Sibawayhi approvingly cites the grammarian Yanus (see Baalbaki 2008: 14) who sanctions the energic -n after hand 'why not?':
(17) halla ta-gal-anna why not you-say-T "why don't you say?" = "you should say"
A general characterization of the meaning of the energic will be given in 2.2.1, below, after further data are introduced, though for the moment it is enough to note that the energic is strongly associated with illocutionarily marked contexts.
Turning to the larger historical issue addressed here, ostensibly there is only a partial overlap between the -n of the Arabic dialects and the energic of Classical Arabic, the only similarity being that they both occur before object suffixes. A closer look at the distribution of the energic in one important Classical Arabic text will help form a more nuanced picture of this initial observation, however.
2.2.1. The energic in the Qur'an
The Qur'an is not only the most important text rendered in Classical Arabic, (13) but also its earliest extended text. Aside from its religious significance, its early pedigree makes it an invaluable source for one variety of early Arabic. It is certainly the case that the Qur'an as we know it today achieved its sanctioned final form only with Ibn Mujahid's (d. 324/936) al-Sabca ft l-qira'at, a work of the early third/tenth century, some three hundred years after Muhammad (see Beck 1945). Nonetheless, its very authenticity would have preserved it from extensive re-editing, so that it may be assumed to be the longest early text available.
For the present study, the online source "The Quranic Arabic Corpus" was used. (14) This allows a concordance search of all verb forms (19,364 tokens, according to the corpus) in the Qur'an, which are listed by lemma according to frequency. Twenty-five of the thirty (web)pages of verbs--the first fifteen and the last ten--were searched for the heavy energic -anna. There are fifty lexical types per page. As is usual in corpora, the most frequent lexical types overproportionally represent the token usage. The first page of fifty types, for instance, accounts for a total of 10,379 tokens, slightly over one-half of all verb tokens. In token terms probably over ninety percent of all Qur'anic verbs are included in the statistics presented here; in lexical type terms, 1,246 verbs, out of a total of 1,496 (83%), were searched. In this group, eighty-three verb types occurred with the heavy -n. Translations are taken from the same website.
It can be noted that the overwhelming majority of energic tokens are the heavy form and therefore only this variety is treated here. The light energic is vanishingly rare. To gauge an idea of the frequency of the light energic, the verbs on the first page and the verbs on the last ten pages of the Quranic Arabic Corpus were exhaustively searched for occurrences. The search turned up only one token, as against eighty-three tokens of the heavy variety on the first page alone. (15)
126.96.36.199. A basic summary of linguistic properties of the energic in the Qur'an
In the sample no tokens of the energic are found attached to the dual or feminine plural. Otherwise it occurs in all three persons, and in the singular and plural. In virtually all contexts it is introduced in the presence of another morpheme that co-marks the "energic" meaning of the construction, as Sibawayhi describes (see above). These will be termed "illocutionary" operators. In the Quri[macron]anic citations examined here these contexts can be summarized in Table. 1, with their token counts.
Table 1. Illocutionary operators co-occurring with the energic -anna Frequency Percent Cumulative Percent la (lam al-qasam) 149 72.3 73.6 la, negative 38 18.6 91.6 imma 'when, if' 9 4.4 96.0 imma ... aw 'either 6 2.9 99.0 ... or' aw 'or' alone 1 .5 99.5 others 1 .5 100.0 Total 204 100.0 Examples of the less frequent morphemes are as follows (see (20) for la-): (18) fa-imma ya-'tiy-anna-kum mi-ni hud-an (Q 2:38) And-when 3-come-T-you-MPL from-me guidance-ACC "And when guidance comes to you from Me" (19) hal yu-dhhib-anna kayd-u-hu ma ya-ghitu (Q 22:15) Q 3M-go-T effort-NOM-his what 3M-enrages "will his effort remove that which enrages" (16)
Not infrequently, there is a sort of energic agreement, when the energic in one verb form is followed by the same in the next or neighboring verb. In Q 4:119, for instance, six tokens of the energic follow one another successively.
(20) wa-la-u-dill-anna-hum wa-la-u-manniy-anna-hum wa la-'-emur-anna-hum
And EM-I-mislead-T-them and EM-I-arouse-T-them and EM-I-order-T-them
fa-la-yu-battik-u-nna adhan-a 1-an'ami wa-la-'-amur-anna-hum
and-EM-3-slit-PL-T ears-ACC DEF-cattle-GEN and EM-I-command-T-them
"And I will mislead them, and I will arouse in them [sinful desires], and I will command them so that they slit the ears of cattle, and I will command them so they change"
Thirty-two of the verses have more than one energic token in them.
The previous example (20) illustrates another relevant point, namely, that the la- of oath often occurs with no explicit oath before it, so that it simply has the effect, with the energic suffix itself, of marking a pragmatic emphasis. (17)
It is also not uncommon for the same phrases to be repeated in more than one verse (see, e.g., Zwettler 1978 on orality and repetition in Arabic literary composition). The phrase (la/ la-) takun-anna min X (18) "(don't) be of the X," for instance, accounts for sixteen of eighteen tokens of the energic on k-w-n, as in Q 6:114, 7:149, and 7:189 respectively:
(21) la takun-anna mina 1-mumtar-ina / l-khasirina / 1-shakirina
"do not be of the doubters" / "of the losers" / "of the grateful"
In other cases, whole clauses recur. The next example, for instance, is found in both Q 31:33 and 35:5:
(22) fa-la ta-ghurr-anna-kumu 1-hayat-u 1-dunya wa-la ya-ghurr-anna-kum bi-llahi 1-ghararu
"so let not the worldly life deceive you and let not the Deceiver deceive you about Allah"
Each such case is counted as an individual token in the statistics below. What applies to phrases applies to a degree to the illocutionary operators and complements as well. The illocutionary operators tend to collocate with individual verbs. The verb hasaba consider', for instance, occurs in two verses--nine tokens in all with the negative la; this one verb therefore accounts for nearly one-quarter of all la + -n occurrences. Also, ard 'show' occurs in four tokens in four verses, all in the context of imma 'if'. The verbs may be predisposed to occur with certain types of complements: adkhala 'make enter', for instance, occurs only with a pronominal object complement in its four tokens.
As far as the meaning of the energic goes, generalizing over the examples and description given so far, it can be said to represent an assertion on the part of the speaker that "the event or state of affairs represented in the predicate will with a high degree of certitude occur." It is, as it were, a certitude operator. In the case of imperatives it expresses a strong desire on the part of the speaker that an event should or should not occur. 19 In actual discourse, based on the Queanic material, the energic always occurs in tandem with another morpheme to co-mark this meaning. (20)
188.8.131.52. A statistical summary
In order to better assess the function of the energic, the syntactic contexts in which it occurs in the sample were further broken down into the following categories (in a few cases the class is so small that no example is given):
1. With pronoun object complement; see (18), (20), and (22), above.
2. With noun clause object complement,
(23) la-ya-qul-anna dhahaba 1-sayyi'-at-u 'ann-i (Q 11:10) EM-3-say-T went DEF-bad-PL-NOM from-me "He will surely say Tad times have left me"
3. With nominal direct object complement immediately following; this class is divided according to whether the direct object bears the definite article,
(24) wa la-ya-'lam-anna 1-munafiq-ina (Q 29:11) and EM-3M-know-T DEF-hypocrites-PL "He will surely know the hypocrites"
4. or an object without definite article, as in (20) fa-la-yu-battik-u-nna adhan-a l-an'ami.
5. With prepositional object immediately following,
(25) la-yu-'min-anna bi-hi (Q 4:159) EM-3-believe-T in-him "he will surely believe in him"
In classes 1-5 the complement immediately following -anna is implied in the argument structure of the verb; amana 'believe' (25), for instance, requires that its object complement be marked by the preposition bi-. In the following four classes the complement immediately following -anna is not implied as a (non-subject) complement of the verb.
6. With non-argument complement (e.g., prepositional phrase) immediately followed by nominal object.
7. With non-argument adjunct immediately following and no nominal object; see (21).
This class is problematic. The majority of its tokens are due to one verb 'be', k-w-n (18 tokens), where the complement introduced by min is considered to be an adjunct, not part of the argument structure of k-w-n. This problem is alluded to briefly below.
8. With overt noun subject immediately following (19).
9. With no overt nominal or prepositional argument of the verb at all,
(26) la ta-mut-u-nna (Q 3:102)
"do not die"
The statistics for these contexts are given in Table 2.
Table 2. Heavy nun in Qur'anic Arabic in nine contexts N 1. pronominal DO 72 2. S Comp 34 3. prep pro DO 18 4. DO with al 6 5. DO without al 25 6. V PP OBJ 1 7. non-argument comp 27 8. V Subject 14 9. no overt argument 7
2.3. The development of the energic in Classical Arabic
To begin this section, three points derivable from the statistics in Tables 1 and 2 relevant to the comparative analysis developed here stand out. The first is that in the overwhelming majority of cases, the complement that follows the energic -anna is (1) tied to the argument structure of the verb and (2) a non-subject complement. Of the 204 cases, 156 (76%) fall into the class of non-subject complements that are implied in the argument structure of the verb. If the min complements of k-w-n are included, the number increases to over eighty percent (see the discussion in 184.108.40.206, above). There are only twenty-one cases where a non-subject clausemate does not occur at all, and only fourteen where a subject argument immediately follows the heavy nun.
The second point is that the largest single category, or thirty-five percent, of verb complements is composed of V + anna + pronoun object; and the third is that the energic in the vast majority of tokens is introduced by la-, the lam al-qasam (see Table 1). I will return to the statistics at the end of this section.
In addition to these three points, CA also has complementizers (section 2.1.2, above) that are of the same form as the heavy -n, -nna, and are of the same parentage as the -n that was grammaticafized in the object marking complex. As has been well studied (in particular, Testen 1998), the complementizer inna is very frequently (anna less so) paired with the emphasizing la-, as in the examples of (27) and (28). (21)
(27) inna zayd-an la-qa'imunl la-ft
"Zayd is indeed standing/at home/getting up"
(28) alla annahum la-yekulana l-tdama (Q 25:20)
"they (were men who) ate food (and walked through the streets)" (cited by Ibn Hisham, 307)
Equally relevant to the current question, Sibawayhi (1970, 1: 421.3) explicitly notes that the inna ... la- construction occurs in the contexts of oaths even without the imperfect verb:
(29) a-shhad-u inna-ka la-dhahib-un I-witness-IND EM-you EM-going-NOM "I bear witness to the fact that you are going"
Two key points are relevant here. First, the la- marks a predicate, either nominal or verbal; and second, in most cases the complementizer innalanna is followed either by an accusative noun or a bound pronoun (in accusative form).
2.3.1. The analogical origin of the energic
With this I turn now to the specific proposal. As background to the following explanation it needs to be stated at this point, as will be explained in greater detail in sections 3 and 4 below, that CA inherited a system from proto-West Semitic that already had a grammatical-ized V + -n + pronoun object construction. This is argued to have been the ancestor of the energic -anna, as is manifested in the statistic that the largest category of energic -anna complements are pronoun objects. Furthermore, it may have inherited a structure that euphonically mimicked the V + -n + pronoun object, but left off the pronoun object. This alternative is discussed in the appendix. Whether CA itself analogically developed the energic from scratch, as it were, or whether it functionalized a previously functionless V + -n structure is not crucial to the immediate analysis. In order to keep the presentation concise, I will assume that the development described below is internal to Arabic.
With this basic background, I follow with a description of the development of the energic in steps (the historical stages will be described in section 4, below) as an inferential account of how speakers would have developed the grammaticalized energic. The statistics presented in Tables 1 and 2 will then be discussed for their compatibility with the proposed reconstruction and the statistics will be expanded to better illuminate the close connection between energic -anna + pronoun object in Classical Arabic.
Step 1. The ancestor of CA inherited the following two structures:
(30) V + -n + pronoun object (inherited structure A)
(31) inna/anna + pronoun/noun + la-predicate (inherited structure B; see 2.1.2)
Inherited structure A is the intrusive -n with pronoun object suffix (see sections 3, 4, and 5) and at the proto-CA stage is not the energic whose development is being described here.
Step 2. Speakers made an analogical (cognitive) association between the structures (30) and (31) in three ways. First, in each case the predicate can be marked by la-. Second, the phonological segment -nna of inna/anna is formally identical to the energic -(a)nna. Moreover, all share the generalized form V nna. Third, the complement of inna/anna is obligatory grammatically, and indeed is often a suffix pronoun, while a non-subject complement of a verb occurs immediately after the verb with a high degree of frequency, often in the form of V + -n + pronoun object. The crucial association is in the obligatory linear linkage of the form V nna- with the following constituent, not with the syntactic function of the following elements, which are often distinctive. The syntactic function of the complement following the emergent energic -anna would have been originally controlled by the pronoun object, i.e., this inherited feature already defines the element after energic -anna as an object--or a non-subject argument--of the verb.
The associations can be represented diagrammatically, as in Figure 1.
Step 3. It is a small analogical step to fill in a full noun after the verbal suffix -anna rather than a pronoun, on the model of Figure 1. The analogical proportion can be stated as follows:
(32) inna-pro obj + /a-Pred : /a-Verb-anna-obj pro:: inna - Noun + la-Pred : la-Verb-anna-X
X is filled in with a noun (or other non-subject complement), and so arises the Classical Arabic energic, which originally was a structure restricted to V + -n + pronoun object generalizing to V + -n + pronoun object or N.
In lockstep with the expanded formal context of V -anna was the development of the meaning of the energic. As seen in the statistics in Table 1, the energic occurs with the greatest frequency--by a considerable margin--with la-, which itself has a general meaning of emphasis, and, as seen in 2.2, Sibawayhi even went so far (too far, I think) as to suggest that imperfect verbs following la- always require the energic. Another context in which the energic occurs is the conditional imma in ma, and conditional clauses can also mark the result clause with the emphasizing la-.
At some point the constant co-occurrence of the (now) energic -anna with emphasis la- as well as with other illocutionary particles led to it acquiring the status of a certitude operator, a meaning anchored strongly enough that by Sibawayhi's day it allowed inferential meanings to be derived, as discussed in (16) and (17) above.
It should be emphasized here that the CA energic in this treatment is interpreted as developing by an analogical extension of an already existing suffix, namely, the Pred + -n + pronoun object (see section 4, below). The energic itself does not form by grammaticalizing or incorporating independent material. (22)
Relating this reconstruction now to the Qur'anic statistics, it should be kept in mind that the Qur'an represents a linguistic stage at which time the energic is fully functionalized grammatically. It does not represent the proto-stage itself of this development.
Argued here is that overall the numbers reflect the origin of the energic in various ways. A central point is that the single largest grammatical category occurring with the energic suffix is the pronominal object, representing over one-third of all tokens. The inference drawn from this is that this accumulation in one category reflects the historically original stage. A second major point is that grammaticalization of -n into the function of an "energic" took place in the context of illocutionarily marked morphemes, foremost among them la-. As Table 1 shows, the energic -anna in the Qur'an almost never occurs outside of an illocutionarily marked context. This context was needed to regularize and generalize the meaning and distribution of the inherited -n suffix.
Furthermore, the reconstruction sees the basic grammar of the energic to be that of marking a non-subject argument. This follows from the analogical workings with the comple-mentizer innalanna, which requires that it be followed by an accusative complement. There are, in fact, occurrences of V-anna + subject in the sample, or simply V-anna + [empty set], but these together represent barely more than ten percent of all tokens (see (19), above). The vast majority of cases are V -anna + (non-subject) complement. Moreover, of the complements, depending on how the complement of k-w-n is interpreted (see (21), above), anywhere between 75 to 85 percent of the complements directly following -anna are integrated in the (non-subject) argument structure of the verb. This again is interpreted as an historical relic, as it were: the energic developed as a general category by integrating non-subject verbal complements into the predicate argument structure, in part under the analogical working of the complementizer innalanna.
To be sure, as it developed into an independent verbal suffix, the energic came to allow any complement to follow it directly, including the subject, or to allow no complement to follow it. As noted, however, the small percentage where this does occur reflects the fact that this is a later development.
The statistics in no way prove the correctness of the proposed historical development. It is reasonable to claim, however, that in their present form they serve as a plausible substantiation of it. Moreover, the statistical argument can be considerably strengthened, as will be shown in the next section.
2.3.2. A comparative statistical perspective
It will have been apparent that the statistical count in Table 2, though very suggestive, still lacks a comparative perspective. In particular it needs to be shown that the distribution of the categories in Table 2 are indeed significantly linked to the energic. It could be that the numbers in Table 2 merely reflect the overall occurrence of these categories in the Qur'an as a whole. To carry the comparison one step further, an arbitrary sample of all tokens of V + object in the first ninety-one verses of the Qur'an (al'Tatilia and al'Baqara) were classified according to the following contexts. This sample serves here as the standard of comparison against which the energic -anna can be measured. Table 3 gives the frequencies of each context as well.
Table 3. V--NP/ProObj in a general sample of Qur'anic Arabic V--pronoun object 72 V--NP 69 V--S comp 80 V--X--NP, X [not equal to] subject 83
These contexts correspond to classes 1-6 in Table 2. Essentially all of the V--non-subject argument tokens were classified, that is, the set which it is argued formed the basis of the development of the energic in CA. It should be noted that the classification departs slightly from the Arabic grammatical tradition in that it includes the three tokens of illa + N-ACC as direct object (V--NP). The "X" in V X NP is usually a prepositional phrase (anzala mina l-sama ma-an "He let rain fall from the sky," Q 2:21). Seven tokens of V--pronoun object--NP (ihdind l-sitata l-mustaqim, "Show us the straight path," Q 1:6) are included only among the V--pronoun object.
The crucial comparison to be applied to these data and those in Table 2 is whether the V--pronoun object has a significantly higher degree of occurrence with the energic than it does with the sample at large. This comparison is carried out by opposing the V--pronoun object to the combined frequencies of all other object complements listed in Table 3.
This comparison strengthens considerably the central claim of a special relationship between the energic and pronoun object. The chi-square distribution in Table 4 shows that the co-occurrence of a pronoun object with the energic is not a randomly related phenomenon, but rather that there is a significant positive association between the two categories of "pronoun object" and "energic." At this point a stronger causal relationship may be made between the occurrence of V-anna-pronoun object and its historical pedigree.
Table 4. Chi-square distribution of ProObj vs. V--NP, -anna vs. general sample general sample -anna (energic) V--pronoun object 72 72 V--NP/S comp 176 83 p < .000 df = 1, Pearson chi-square = 12.6
The statistics in Table 3 are suggestive in another way, namely, that the complement NP in the energic (see Table 2, category 6) is separated linearly from the verb to a far lower degree than in the sample at large. This is the V X NP construction, which occurs only once with the energic, thirty-three times in the overall sample. Recall that in Figure 1 it is argued that an important part of the analogical development depended on the argument occurring immediately post -anna in the energic.
To summarize this subsection, adding a comparative statistical sample reveals that the analogical associations argued to be criterial in the development of the energic are indeed of a significantly higher degree of frequency in the energic than in a sample of the Qur'an as a whole. Observed statistical correlations are argued to replicate the historical development of the structure.
2.3.3. Classical Arabic as an innovative variety
This reconstruction of the energic interprets the fully functional grammaticalized structure as an innovation of Classical Arabic, even if it developed out of pre-existing building blocks, as described in the following three sections. The analysis is thus to be compared to Retso's analysis of the dual in Classical Arabic. He describes a process whereby the dual category in Classical Arabic innovated out of proto-Semitic markers of countable plurals and plurals of paucity (1995: 189). In both cases, the development of the dual as a grammatical category and the development of the energic -anna as an illocutionary force marker of speaker certitude are innovations in Semitic that set the classical variety of Arabic off from all others.
3. THE INTRUSIVE -N IN OTHER WEST SEMITIC LANGUAGES
In this section the major distributions of -n in other West Semitic languages will be summarized, in preparation for an overall construction that will be developed in section 4.
In West Gurage (Outer Southern Ethiopian Semitic) -nn occurs before object suffixes in past tense verbs. In other contexts ("elsewhere," in Hetzron 1977: 65) the pronoun object occurs without the -nna-, as segmented in the paradigm below.
(33) SG PL 1 -e -e 2M nna-xa nna-xamw 2F nna-hy nna-xama 3M nn-23 mm-amw 3F nn-a nn-ama abli-nna "he gave her" The -nn is always geminate in Muher.
There are no complete paradigms for Biblical Aramaic (Rosenthal 1961: 71). (24) A vowel occurs before the n when the object is suffixed to a C-final verb. Franz Rosenthal (1961: 54) suggests the underlying form is (i)nn-, reduced to -Vn when the "-n is vowelless." In Biblical Aramaic the -n is not obligatory before an object suffix. (25)
(34) SG PL 1 -nna-ni nna-ni 2M -nna-k n-koon 2F 3M nne-h 3F nn-ah
It rather appears that in Biblical Aramaic the -n is variable, in the sense applied in Table 5 below.
For Old Aramaic, Rainer Degen (1969: 80) shows an -n only in imperfects, and only before a suffix, e.g., yqtl-n-h "he will kill him," and Jacob Barth (1907: 3) notes that an -n is attested in the Elephantine (Egypt) papyri.
For Jewish Palestinian Arabic, ca. seventh century C.E., Dalman notes that in the perfect and active participle the -n occasionally occurs as well before an object suffix (1905: 360, 380). The -n is obligatory in the imperfect before an object suffix.
(35) SG PL 1 nnd-ni innd-na 2M inna-koon inne-koon 2F innii-k 3M innee-h inn-uun 3F inna-h inn-iin y-qdm-in-k "he precedes you (SG)"
Writing on Samaritan Aramaic attested between 400-1000 C.E., Rudolph Macuch (1982: 132) gives a similar paradigm, which will therefore not be given here, except that in Samaritan Aramaic the insertion of the -n is optional. As with Jewish Palestinian Aramaic, although more frequent with the imperfect verb, it also occurs with the perfect and with the infinitive.
Further to older varieties of Aramaic, Barth (1907: 3), citing Theodor Noldeke, notes that Mandaic has -in before plural object suffixes and Babylonian Talmudic -in-hoon. He further observes that suffixing the -n on perfect verbs is not infrequent in both Western and Eastern Middle Aramaic (Jewish Palestinian, as noted above, Mandaic, and Babylonian Talmudic). Barth concludes from these observations that the -n in these varieties is basic to both imperfect and perfect verbs (1907: 8).
In Modern Western Aramaic, spoken in three villages in Syria, the -n has been completely regularized in the subjunctive paradigm, whereby before any object suffix, -n is inserted, with -nn occurring before -V-initial pronouns, -n before -C-initial (Arnold 1990: 208-12). The vowel before -n varies according to grammatical and dialectological parameters. The subjunctive corresponds to the Arabic imperfect.
(36) itSan 'he carries' SG PL 1 y-tu Sn-in-ni y-tu Sn-en-nah 2M y-tu Sn-enn-ax y-tu Sn-en-xun 2F y-tu Sn-inn-is y-tu Sn-en-nen 3M y-tu Sn-enn-e y-tu Sn-enn-n(un) 3F y-tu Sn-enn-en
For the past verb the situation is dialectally complicated, even if Modern Western Aramaic is only spoken in three villages. An -an (-ann-V) is inserted before plural object suffixes in all three villages (sim S-an-xun "he heard you [MPL]"). Before a singular pronoun object, no -n is inserted in the village of Jubb(adin, while an -n is inserted before the pronoun suffix in Balch% (raham-n-i "he loved him"), and in Md 'loula it is inserted only after CCiC-stem verbs (Arnold 1990: 201-2).
Biblical Hebrew has the following (Meyer 1972: 218; Hasselbach 2006: 317, citing preassimilation forms (26)):
(37) SG PL 1 yiqtal-in-ni naqtul-an-na 2M yiqtal-in-ka 2F 3M yiqtul-in-huu 3F yiqtul-in-ha
Ronald Williams (1972: 84) also notes four tokens of the -n suffixed to Hebrew participles. Finally, it should be noted that Williams (1972: 84) lists a number of examples of the -n on verbs without pronoun objects in Phoenecian, Old and Biblical Aramaic, and Ugaritic (see n. 24, above). A number of these are on 2FSG and 2/3MPL or dual prefix verb forms, which Williams associates with the energic -n, but which other scholars (e.g., Degen, Rosenthal), whom this article follows, associate with the indicative. Discounting these, there remain only a small number of tokens for V + -n in Phoenecian and Ugaritic with no object suffix. For Ugaritic Testen (1993: 296-97) adds that -n is "frequently employed ... before object suffixes." Both Williams and Testen point out that it is difficult to associate the -n with a dedicated grammatical function.
What can be termed the "otiose -n," that is, an -n that cannot be, or at least to date has not been identified with a discrete grammatical function, will not be integrated into the discussion, though the basis for doing so is discussed in the appendix.
Before explaining the reconstruction developed here, the following overview of the situation in West Semitic will be helpful. Table 5 summarizes the data in sections 2 and 3 under three categories: obligatory, optional, or variable, each in those contexts that have been defined above where the intrusive -n occurs. The -n is obligatory, for instance, before a suffix pronoun after an AP in various Arabic dialects, whereas in the same morphological context it is variable in Omani Arabic after weak verbs. It is variable in Samaritan Aramaic, object suffixes with -n alternating with those without. Optional means that the choice of -n depends on the choice of the energic (in CA) according to discourse immanent factors.
Leaving aside Ugaritic (see appendix), it is immediately striking that in all but Classical Arabic the decisive factor governing the occurrence of -n is the presence of a suffixed pronoun object. This observation forms the basis of the reconstruction. Given that identical or near identical forms of wider distribution point to a common origin, the following can be reconstructed for proto-West Semitic, -n originally occurring only before a pronoun object suffix:
(38) *V-n + pronoun object suffix
This reconstruction leaves Classical Arabic as the odd man out, where -n occurs both before a pronoun object suffix and before a noun. However, it can be plausibly postulated that the ancestor of Classical Arabic also had the basic distribution given in (38). The statistics in Tables 2, 3, and 4, above, speak to this point. They indicate that in Qur'anic Arabic, V -anna, which is the context illustrated in (18) and (20), is associated with the suffixed pronoun object. This is inferred to reflect the original situation in ancestral Classical Arabic. The development of the energic in Classical Arabic, as described in section 2.3, above, then falls into the following stage 4.
The historical derivation leading to (38) proceeds in five stages as follows:
Stage 1. Originally proto-West Semitic (PWS) had a construction with -n + noun or pronominal complement which, following Barth (1907: 8) and Chaim Rabin (1951: 38), was a proto-Semitic emphasis particle, deictic or perhaps presentative.
Stage 2. This developed into a complernentizer, marking one of the main arguments of a clause, the topic, subject, or object. Here a first stage of grammaticalization can be hypothesized in which the -n + pronoun/noun is drawn into the argument structure of the predicate.
Stage 3. Still in the PWS stage (i.e., without naming any individual languages) its functions split. In Stage 3a, in its specifically object marking function, it became restricted to pronoun objects and underwent a further stage of grammaticalization, becoming morphologically bound to the verb as a predicate suffix: Pred + n + pronoun object. At the same time, Stage 3b, the complementizer -n continued in the ancestor of some of the varieties of Semitic. "Predicate" is used here as a term that subsumes both verb and active participle. As a predicate suffix it loses its relation to nominal complements. This point is expanded upon from a comparative perspective in section 5.1, below.
It is relevant to note that the grammaticalization of -n occurred in some but not all ancestral PWS populations, so that at this point two ways of representing the predicate + pronoun object came into existence.
Stage 4. In this stage, when distributions in specific languages began to take shape, the complex -n + pronoun object grammaticalized as a suffix on a predicate (stage 3) and it undertook its journey in one complex of West Semitic (viz., Aramaic, Arabic, Ethiopian Semitic). In addition, a further development took place in the immediate ancestor of Classical Arabic. On the one hand, along with other West Semitic languages/varieties Classical Arabic would have participated in the grammaticalization of the -n + pronoun object suffix on the verb. At the same time, analogy to the independent complementizer innalanna--indicated by the dotted line in Figure 2--which itself occurs not only with a pronominal suffix, but also with a noun complement, allowed the -n verbal suffix to develop into a generalized suffix occurring with either pronoun or nominal complement, that is, into the energic. This is the process described in sections 2.2 and 2.3, above.
Stage 5. This stage is essentially the situation that is found today, to the extent that the languages are still spoken.
The entire development of proto-West Semitic *-Vn can be represented with a tree diagram (Figure 2).
To recapitulate the current solution relative to the two main approaches sketched in the introduction, the origin of -n on verbs (and predicates) lies in a grammaticalized, originally independent presentative/complementizer morpheme, etymologically related, inter alia, to the Arabic complementizer inna/anna and her sisters. The original grammaticalization as a predicate suffix occurred not in the form of verb + -n, but rather verb + -n + pronoun object. This conclusion is consonant with Barth (1907: 8), Landberg (1909: 738), and Retso (1988: 88), but runs counter to the more recent Testen (1993: 303) and Hasselbach (2006), who see the -n as an original verbal suffix. In West Semitic, the only grammaticalized attestation of a reflex of the -n as a verbal suffix operating independently of pronoun object suffixes is the Classical Arabic energic, which is interpreted here as an innovation in Classical Arabic. (27) On the other hand, with Barth, Landberg, Testen, and (probably) Hasselbach, all occurrences of -n + pronoun object suffix are derived from a common source in West Semitic.
5. INDIVIDUAL ISSUES
5.1. Grammaticalization of -n + pronoun object to verb suffix (stage 3a)
This analysis assumes that the original presentative came to mark the object of a verb. Subsequently the complex -n + pronoun object grammaticalized to the verb, forming a phonological word with the verb. That the -n is a part of the verb or AP phonologically can be seen in the stress placement. In the Arabic dialects, for instance, stress falls on the -VVn or -V nn in the syllable closed by the intrusive -n, according to a fairly universal of stress placement in spoken Arabic, which has stress falling on the first heavy syllable from the end of the word (katb-inn-a, etc.); see examples in (1) through (8). Similarly in Hebrew the word stress falls on the syllable before the -nn.
The grammaticalization of originally independent morphemes as verbal suffixes is not unusual in Semitic, nor was -n alone among proto-complementizers. In Hebrew the independent complementizer oot/et- (the nota accusativi) occurs as a marker not only of definite accusative objects, but also as a topic marker. In Palestinian Jewish Aramaic the same etymological -it + pronoun object, also realized as -t + pronoun object, can be suffixed to a perfect verb as an object marker (qtl-it-wn "he killed them," Dalman 1905: 360). Whereas in Hebrew oot/et--does not suffix or cliticize to the verb, Noldeke (1893: 104) sees Palestinian Jewish Aramaic as having taken this step in the etymologically identical morpheme. Rudolph Macuch (1983: 133) describes a nearly identical construction in Samaritan Aramaic, which he explicitly links etymologically to the nota accusativi, and Noldeke himself cites one example of V-t-Obj pro as early as the Zenjirli (Aramaic) inscriptions.
In Classical Arabic the particle iyya + pronominal object, which later itself became a verbal clitic (see (39), below), can mark an independent pronoun object, as in Q 1:4 (Sibawayhi 1970, 1: 336, 346), though unlike inna and oot-, iyya cannot mark a topic or subject. However, a perhaps related morpheme surfaces in ayyuha, which marks a vocative subject. These morphemes reflect what was characterized in stage 2 above as stage 1. grammaticalization of -n, or in the case of Palestinian Jewish Aramaic, variably stage 3.
Looking further at results of grammaticalization of verbal complements on contemporary varieties, in Arabic the indirect object marker -1+ pronoun object occurs as a verbal/AP suffix in most (though not all) modern Arabic dialects, as in the following example from Najdi Arabic (Ingham 1994: 30, see Wilmsen 2010):
(39) gil-t-li-hum-iyyda-h (28)
"I said it to them"
The same example shows a slightly less widespread development, namely, the suffixation of the object marker + pronoun object marker iyya to the verb. The entire complex of modern
Aramaic varieties is based to one degree or another on participles refunctionalized as finite verbs. All of the finite tenses of the neo-Eastern Aramaic languages, except Mandaic in one case, are based on historical participles. In what is termed the ergative, based on an older passive participle (pafiyl), the subject is marked by the complex -1 + pronoun object, the -1 cognate with the Arabic preposition for instance, gris-1-i "I pulled him" (literally, "he is pulled to me," Jastrow 1997: 362). These cases are summarized in (40).
(40) Grammaticalization of originally independent morpheme + pronoun object as suffixes on predicates
predicate + -/ + pronoun object = indirect object (ancestor of many Arabic dialects)
predicate + -iyyd + pronoun object = object (ancestor of many Arabic dialects)
predicate + -(i)t + pronoun object = object (Jewish Palestinian/Samaritan Aramaic)
predicate + -/ + pronoun object = subject (neo-Eastern Aramaic)
Thus, it is a common development in Semitic for pronoun objects on an independently occurring X + pronoun object (X = iyya, -l, -it) to become grammaticalized as a unit as a predicate suffix, marking the subject or object of the verb. The complex -n + pronoun object was among the first of these: predicate + -n + pronoun object = object (proto-West Semitic, stage 3).
5.2. Vowel quality and length of -n
As noted at the beginning of the article, the question of both vowel quality and length of -n has been an issue in the interpretation of this morpheme. Looking at the issue from an Arabic perspective, the vowel quality is probably an indeterminate one. (29) On the basis of the modern dialects I argue in A Linguistic History of Arabic (2009 (2): chapter 2) that the fundamental contrast among short vowels is between the High and Low varieties, i.e., a bi-valued contrast, with the difference between u and i phonologically determined, and that in the Old Arabic grammatical literature as well substantiating arguments can be made in this direction. As for the contrast between a and i (eponymic for a high vowel), there is a systematic and not completely regular variation recognized in the Arabic tradition (known as taltala) between a and i in grammatical morphemes. The list of these morphemes includes the following (in parentheses is given a typical dialect where the form occurs, not an exhaustive listing of their geographical distributions):
(41) a-form i-form Preformative vowel ya-lbas (ELA) yi-lbas (Cairene) (30) Definite article al- (Sudanic) ii- (North African) 3FSG on verb -at, shaaf-at -it, shaaf-it (Cairene) (Sudanic) FPL on verb -an (southern -in (Najdi) Iraq) FSG on noun -at (Gulf -it (Nigerian) dialects) Intrusive--Vn -an (Shukriyya), -in (as above) CA energic (31)
Pending a comprehensive integration of these data into comparative Semitic,32 the possibility needs to be left open that the ancestral variety of the intrusive -n had both a high and a low vowel variant, -in ~ -an. In no language is -V suffixed to a -V-final morpheme (e.g., 3MPL).
With regard to consonantal length, single and geminate -n are generally in complementary distribution in West Semitic, nn- occurring before -V, -n before -C. (33) This is unequivocally the case in most contemporary Arabic dialects that have the morpheme (see (1) through (7), above), which contemporary observers (Reichmuth 1983; Holes 1987; Owens 1993; Ingham 2006) have confirmed. (34) This also applies in Modern Western Aramaic, as noted by Arnold, and it is explicitly stated as the distribution by Rosenthal for Biblical Aramaic (see above). (35) In Gurage--and it would appear also in Jewish Palestinian Arabic--the form nna- has become general, this also being the form of the heavy -n in Arabic, and in Yemen Landberg's data suggest a fixed -nn, except -n before another -n (see (1), above). In general in West Semitic it would appear therefore necessary to posit only -*n, with two conditioned allomorphs, and -nn, with perhaps -nn generalizing in some cases. (36)
As far as the light and heavy -n of Classical Arabic is concerned, the following observations can be offered here. The heavy -n does in fact fit the expected profile of a geminate form when it is recalled that its prototypical occurrence is in a connected, non-pausal (wasl) position. A connected position in Arabic requires either a systematic (e.g., case) vowel or an epenthetic i. Assuming the final -a to be a reflex of its typical morpho-syntactic position, the doubling of the heavy -nn adheres to (42a). Note that in the Qur'anic sample, only six tokens of -anna occur before no clausemate at all (Table 2).
On the other hand, it is not inversely possible simply to link the light -n to (42b), since -C initial object suffixes occur with both heavy and light -n. In this regard it is relevant to return to Sibawayhi's observations of the phenomenon. The light -n, as noted above, does not occur before CC, in this respect obeying standard syllable structure constraints in Arabic, as long as no epenthetic vowel is insertable, or, as Sibawayhi notes, no shift to -a is possible. The light and heavy -n then overlap only before an object suffix and before a word beginning with -CV. Otherwise they are complementary, with the heavy -n occurring before CC as well as in the dual and FPL. Two developments can be imagined here.
The ancestor of Classical Arabic could originally have obeyed (42), with the light -n occurring only before object suffixes and perhaps prepausally, and the heavy -n occurring in wasl position before word boundary. The distribution of the heavy -n then generalized to prepronominal position, and the light -n to inter-word contexts. Alternatively, one can imagine that the light -n was the original form, with an -nn variant arising with the development of non-pausal morphophonology. A similar development may have occurred with the sister particles, inna, anna, lakinna developing as non-pausal variants of in, an, and lakin. In both cases, the non-pausal form became the statistically common one. In both scenarios, however, the still unwritten historical linguistics of pausal forms becomes crucial. By the same token, the drastic differences in the frequencies of the two forms (see 2.2, above) would seem to imply an historical dynamic between the two, with the heavy -n growing at the expense of the light.
5.3. Multiple populations, multiple historical variants
It is perhaps disquieting that a high degree of variation can be observed in the distribution and use of -n. There are two types of variation. On the one hand, there is a categorical variation--either -n occurs in a variety or it does not (see Table 6, following page, for a non-exhaustive summary). The other type of variation is that noted in column 3 of Table 5, where one and the same population uses the -n variably.
Table 5. Status of -n in nine languages/varieties Obligatory Optional Variable Arabic dialects, AP + Obj pro Omani, (weak) final + Obj pro verbs Classical Arabic, + imperfect V enerl, Tic Gurage + Obj pro Biblical Aramaic + Obj pro Jewish Palestinian + Obj pro Samaritan Aramaic + Obj pro Modern Western + Obj pro Aramaic Biblical Hebrew + Obj pro Ugaritic + Obj pro + verb (otiose -n) Table 6. Occurrence of intrusive -n according to languages/varieties *-n occurs does not occur Classical Arabic E. Arabian peninsula, Khorasan, all other Arabic dialects Uzbekistan, Bagirmi Biblical Aramaic, Jewish Samaritan Syriac, Nabataean Palestinian, Modern West Aramaic Neo-NE Aramaic Gurage-Gunnan GaSez, Amharic, etc. Biblical Hebrew
To help understand the nature of the forces that could have sanctioned such widespread variation, it is helpful to look in greater detail at contemporary Arabic dialects for which we have relatively detailed data and which can still be observed today. (37)
Relying only on dialectal information, the occurrence of the -n would have the following history. It can be tracked to the Southwest Arabian peninsula in pre-Islamic times. The situation reported by Landberg (see (1), above) can be taken as representing the original state of affairs as well. From the southwest part of the peninsula populations moved into the southeast, or what is now Oman, where Reinhardt's representation maintains basically the same situation. From there populations moved up the coast, to the present Emirates and Bahrain. In Islamic times tribes from this area moved into 'Iraq, inter alia into the new garrison cities of Basra and KM, whence the Arab armies moved eastward into Khurasan and what is now Uzbekistan, two areas where the -n is still attested. In the west the movement was into Upper Egypt and then into the Lake Chad area and that of Bagirmi speakers, as will be sketched below. Each migration brought the speakers into contact with other populations, producing leveling of various types, particularly in the post-Islamic era. In Bagirmi Arabic all gender and number suffixes are neutralized with the suffixation of -in. (38) In present-day Bahrain and Khorasan it is reported only in the singular, and in Uzbekistan also only suffixed to a (historical) singular AP. That the form was probably once more widespread in the Arabic-speaking world is suggested by the odd occurrence of -an in Shukriyya and in the non-Bagirmi dialect of Nigerian Arabic after the FPL -aat of the active participle, before object suffixes, as well as Wetzstein's report of its optionality (it has not been attested in contemporary Syrian dialects).
This summary covers nearly 2,000 years of history and a geographical area spanning one-seventh of the circumference of the globe. The first point the illustration thus testifies to is the ability of the same form to be maintained and spread over time and space. Looking at the social mechanism of the maintenance and spread, the summary highlights a second point, namely, the role of small, compact groups (according to Holes, the links between -n speakers in Oman and those in Bahrain can be traced in some cases to historical relations between individual villages). In the Sudanic region the basic Arab demographic social unit was, and to some extent still is today, a nomadic cattle group linked by kinship to sedentary, agricultural villages (Braukamper 1994). Although nomadism is more pronounced in the non-Bagirmi dialect area, this social configuration is characteristic of all Arabs in Nigeria, Cameroon, and western Chad. Even the Arab-Islamic expansion, which brought Arabs into central Asia in the second/eighth century, was effected by relatively small armies of never more than 20,000 soldiers and usually far fewer.
So long as these social units remain intact, it may be assumed, so too will the dialect they use. By definition, neighboring dialects will maintain a different form. What vagaries of social history led one group to originally adopt one form rather than another remains one of the great challenges to understanding language change. (39)
Using the Arabic situation as a basic analogy, projecting back into the era of history when communication was incomparably slower than it is today, it is not at all difficult to conceptualize the spread and maintenance of contrastive forms from a common source throughout large language groups, like Arabic, and, by extension, Aramaic, Hebrew, and peripheral South Ethiopic. In fact, the case of -n in the rather isolated Gurage-Gannan group can be compared to the isolated occurrence of -n in Bagirmi Arabic in the Lake Chad region, and to the ultimate linguistic enclave of Uzbekistani Arabic in the east. The complex distribution of the -n in the preterite in the three villages where modern western Aramaic is spoken, described in (34), above, is further witness to the capacity for significant linguistic differences to be maintained among very closely related linguistic populations. The analogy carries back into early West Semitic, reconstructing the two paradigms, V + pronoun object and V + -n + pronoun object, into different groups that transmitted them ultimately into the present.
A demographic perspective would also appear applicable to the issue--not addressed in this article--why the -n became established only in the Gurage group on the perfect (suffix) conjugation. It would, however, be speculative, lacking better comparative data, to say that in one case the group adopting the -n on the perfect verb was large enough to maintain it into Ethiopic.
5.4 The intrusive -n and the active participle in dialectal Arabic
The limitation of -n to the AP in most varieties of Arabic deserves special consideration. Except for the geographically limited occurrence on the imperfect verb in Oman and Yemen, the -n occurs only on the AP in Arabic dialects. In the reconstruction offered, the -n was originally an optional element on predicates. As has been shown (see, e.g., Ingham 1994: chapter 8; Caubet 1993, 2: 221-46; Eisele 1999: chapter 7), the active participle among Arabic dialects has a uniform function which can broadly be termed "perfectivity" or "current relevance of the action/state of affairs represented in the predicate." It clearly commutes in the verbal system with the perfect and imperfect verbs. The occurrence is so consistent throughout nearly every variety of spoken Arabic (40) that its function as a predicate can be plausibly reconstructed into proto-Arabic. (41)
In Figure 2 above, the -n would have been present in Arabic at least by stage 3, differentiating at stage 4 into V + -n + pronoun object and V -n + noun. Here the attestations of -n + suffix in contemporary Omani Arabic (section 2.1.1, above) are of interest. One explanation for these is as a partial "relic," shadowing the distribution of -n in CA, namely, V + -n + pronoun object. Whereas it was lost, or simply never actualized in most dialects, it became established in the Arabic varieties of Yemen and later Oman and the lower Gulf. As the AP was incorporated into the verbal system, the -n was transferred from the imperfect verb to the AP as well; whereas the -n was largely filtered out from the verb in the course of time and contact, it was protected in the context of the AP, and, comparatively speaking, maintained until today in those varieties where it occurs.
There is mutual accommodation among different dialect groups (42) and in cases of contact, the V + -n + pronoun object construction, being of high frequency (43) and therefore more susceptible to analogical comparison by speakers, either has disappeared or is disappearing from dialects where it ancestrally existed. In the less-frequent AP + pronoun object, on the other hand, devising an analogy is rendered more difficult by two factors: the far lower degree of discourse frequency of the AP generally in Arabic (see n. 43, below) and the difficulty of establishing a morphemic value for the -n, which exacerbates establishing an analogical link to non-n dialect forms.
It can also be mentioned in passing that it is an intriguing coincidence that among all varieties of Semitic, it is in Aramaic, particularly among its western varieties, where -n is most strongly established. This is intriguing because Aramaic, far more than Arabic, has grammaticalized the active (and passive) participle to assume verbal functions (see (40), above). Furthermore, -n is reported as occasionally occurring on Middle Aramaic participles (see (35), above). It should not therefore be ruled out that pre-Islamic contact between Aramaic and Arabic speakers had a role to play in the development of -n in Arabic, i.e., that the -n on the Arabic AP was not a development independent of the Aramaic.
Finally, it is relevant to note that at least one citation can be found suggesting that the "transfer" (if it was that) of -anna from V to AP occurred even in CA. The thirteenth-century grammarian Radi 1-Din al-Astarabadhi (d. ca. 686 A.H.) ([n.d.], 2: 404), basing himself on Sibawayhi, (44) says that the nun can occur on the AP in cases of poetic necessity:
(43) a qa'il-anna ahdar-u al-shuhud-a Q saying-N brought-they def-witnesses "Do you indeed say they brought the witnesses"
The major argument against the possibility that in fact the situation in Classical Arabic is the original one and that what is found in the dialects are relic cases going back to the energic is that the dialects agree with other varieties of West Semitic, against CA, in maintaining (on a variable basis) the -n + pronoun object (Table 5). Standard reconstruction practice thus argues for a common proto-function among these. Assuming the CA situation as the original one, one would have to explain the loss of the energic in all other varieties of West Semitic. If one ignores CA completely, the situation in the Arabic dialects is entirely in conformity with other West Semitic languages, particularly with Aramaic. Note that the loss of the intrusive -n on verbs in most Arabic dialects is explained above in this section as a result of contact with dialects that never had the intrusive -n, and does not rely on a direct derivation from CA. From the reverse perspective, section 2.3, above, offers a concise account of the development of the energic, which allows the mechanisms of its singular innovation to be specified. Thus, the argument advanced here corresponds to standard historical linguistic methodology. There is also a third argument, namely, that if the energic grammatical function had been dominantly widespread throughout the Arabic-speaking community, one might expect some residual evidence of it in most of today's varieties, which is not the case.
In Akkadian it is assumed that the ventive (motion towards speaker) verbal suffix, either -am (occurring after VC-) or -nim (occurring after VV-), is cognate with the West Semitic -n. Wolfram von Soden (1952: 109) noted that "many verbs in ventive form occur only with pronominal objects"--a situation that in general resembles that described in this paper, and which specifically resembles the distribution of the energic in Qur'anic Arabic (see Table 2)--and that "a connotational difference between the indicative and ventive is not easy to establish." Furthermore, in Babylonian the 1SG pronoun object -ni is nearly always suffixed to the ventive suffix, neutralizing the contrast between ventive and indicative altogether. Again, this case resembles another situation described above, namely, the instance of obligatory occurrence of the -n before any object suffix. On this basis it is a priori realistic to push the -n + pronoun object origin back into proto-Semitic itself, as indeed Hasselbach, Testen, and others, albeit with the different analytical perspective noted above, have done. This paper leaves this important issue for further development. (45)
5.6. Against massive independent development: one morpheme, multiple populations
In this final section I would like to take up briefly the question of independent development of the intrusive -n in different varieties of Semitic. As noted above, Retso sees the -n + pronoun object construction developing in multiple ways. The -n in Bagirmi Arabic presents a problem, however. We know with certainty that the ancestral Arabic population reached the Lake Chad area around 1400 C.E. and that they were settled in Upper Egypt and the immediately adjacent area of northern Sudan before then. Whether the -n bearing population came from Yemen or from eastern Arabia (Qays) is at this point unclear. General historical records attest to early, pre-ninth-century migrations into Upper Egypt from both areas (see Garcin 1976; Owens 2003), and, as Holes (2011) shows, the -n would have been present in both Arabian peninsular populations at this time.
Assuming independent development, the source of the in Bagirmi Arabic is problematic. Following Retso one would look to deictic elements. However, there is no deictic model with -n + suffix in Bagirmi, or Western Sudanic Arabic in general for that matter. To the extent that a complementizer etymologically related to inna, etc. (see 2.1.2, above) occurs at all in these varieties, it is via borrowing from Modern Standard Arabic. On the basis of the unavailability of a model that could be grammaticalized into the AP, as well as the general unlikelihood that such a specific morpheme would arise independently in different varieties of Arabic, it is clear that this is an old feature in Bagirmi Arabic, which is inherited from the same stock that contemporary speech in Oman, Yemen, and elsewhere is related to. A similar point ostensibly applies to Gurage (see (30), above), where it is clear that there is no obvious deictic element in the language that could have served as the basis of the -nna pre-object morpheme. (46)
This does not rule out independent development in principle, of course, even if by Occam's razor it is the marked solution. An -n (inna, anna, etc.) or -n + pronoun object marking a complement has been maintained in Semitic-speaking populations over the entire history of West Semitic, as is indicated briefly in section 2.1.2, above. It is also briefly suggested in 2.1.2 that this independent -n itself may have split into different functions, for instance, the two illustrated in (11) for Yemeni Arabic. However, showing that the -n developed independently in different languages, rather than from a common source as argued for here, requires specifying the eras and varieties where this happened, as well as arguing against the unmarked assumption of common origin.
The criticisms of Retso aside, the current article follows in his footsteps in two important ways. First, it follows his model (1988; 1995) in defining certain structures of Classical Arabic--in this case the energic--as innovative relative to other varieties of Semitic. In this historical, comparative perspective the contemporary reflexes of -n in Arabic are maintained from proto-Semitic itself, and are not an accidental relic left from the Classical Arabic energic. And second, his conclusions (1988: 93) regarding the remarkable maintenance of the basic -n + pronoun object construction across thousands of years of chronology and thousands of miles of geography, across different languages and varieties, cannot be endorsed strongly enough. Understanding Semitic language history requires giving due consideration to all varieties, wherever they are at present spoken or were in the past. Reflecting on this vast geo-diachronic landscape leads to the consideration of central issues in the nature of language maintenance, spread, and change. In this context the study of the living Semitic varieties assumes a larger importance than has traditionally been accorded them in historical linguistic study, not only for their intrinsic linguistic value, but also for the fact that social and demographic dynamics affecting the languages can still be observed against actual linguistic usage (section 5.3, above). In the present case, the intrusive -n before pronoun objects in varieties as disparate as the Arabic of Nigeria, Uzbekistan, and the eastern Arabian peninsula, as well as neo-Western Aramaic, should be seen less as a quaint relic attested in "peripheral" varieties as evidence of its origin in an -n + pronoun object complex. What is maintained over space can be a reflection of its original genesis.
APPENDIX: OTHER V + -N + NOUN IN WEST SEMITIC
While there are, as noted in section 3, above, other West Semitic languages with V + -n occurring before a noun object or occurring with no complement at all, the observation remains that among the West Semitic languages only in Classical Arabic is there evidence for the reflex of intrusive -n functioning in a definable grammatical manner without pronoun objects. A good case in point is Ugaritic, which has been exhaustively described by Josef Tropper (2000: 497-504). Tropper lists what he calls the forty-five tokens of V-n in Ugaritic, without a pronominal suffix. He terms this -n the "Energikus I," but he equally notes that it occurs with all verb forms and cannot be associated with a given mode value. From this it may be inferred--Tropper does not say so explicitly--that the V -n cannot be assigned a discrete function, an interpretation that agrees, albeit in differing detail, with Daniel Sivan (1997: 105). Tropper also distinguishes what he terms an "Energikus II," which only occurs before an object suffix. He furthermore assumes this to have been the original suffix, with the Energikus I being derived from Energikus II ("Die Form des Energ. I durfte ... sekundaren Ursprungs," p. 502). In broad terms this agrees with the analysis proposed here, in which the V + -n + pronoun object represents the original West Semitic state of affairs.
Should an assumption of cognate status in all relevant V -n forms be made for other West Semitic languages, then the process described in Figure 2 for Classical Arabic would need to be pushed back to an earlier stage and include ancestral varieties of West Semitic languages other than Classical Arabic. While developing the reconstruction based on this assumption would require a reworking of the details presented here, as well as the incorporation of others, the argument for the priority of V + -n + pronoun object would remain. The sequence would thus be:
1. independent complex: n + pronoun object;
2. grammaticalization V+ -n + pronoun object;
3. V+ -n generalized irregularly to all objects in some languages.
There are two arguments for this sequence. First, step 2 has robust congeners all the way to the present and is comparatively well attested in Aramaic, Arabic, and Ethiopian Semitic, whereas step 3 is completely lost today. One way (circular but plausible) to account for this difference is to assume that step 2 established itself earlier than step 3, and had thereby developed a wider population spread, which ensured a more robust transmission. The second argument is that if V + -n is indeed a stage before CA, then it is functionless, so far as the data in PWS outside Arabic show. No specific, consistent grammatical function can be attributed to it. This implies variable occurrence in the categories of Table 5. This is opposed to V + -n + pronoun object, where -n is clearly an object marker. In this perspective, V + -n can be seen as developing originally as a euphonic mimic of V + -n + pronoun object, establishing a grammaticalized functional niche for itself in the West Semitic branch only in Classical Arabic.
'Antara. n.d. Diwan 'Antara Ibn Shaddad. Ed. Yusuf 'Id. Beirut: Dar al-Jil.
Arnold, Werner. 1990. Das Neuwestaramaische, vol. 5: Grammatik. Wiesbaden: Harrassowitz.
__, and Hartmut Bobzin, eds. 2002. "Sprich doch mit deinen Knechten aramaisch, wir verstehen es!": 60 Beitrage zur Semitistik. Festschrift fur Otto Jastrow zum 60. Geburtstag. Wiesbaden: Harrassowitz.
al-Astarabadhi, Radi l-Din. n.d. Sharh al-Kafiya fi l-nahw. Beirut: Dar al-Kutub a1-'Ilmiyya.
Baalbaki, Ramzi. 2008. The Legacy of the Kitab: Sibawayhi's Analytical Methods within the Context of Arabic Grammatical Theory. Leiden: Brill.
Barth, Jacob. 1907-1911 (repr. 1972). Sprachwissenschaftliche Untersuchungen zum Semitischen. Amsterdam: Oriental Press.
Beck, Edmund. 1945. Der 'Uthmanische Kodex in der Koranlesung des zweiten Jahrhunderts. Orien-talia n.s. 14: 355-73.
Braukamper, Ulrich. 1994. Notes on the Origin of Baggara Arab Culture with Special Reference to the Shuwa. Sprache und Geschichte in Afrika 14: 13-46.
Caubet, Dominique. 1993. L'Arabe marocain, vol. 2. Paris: Peeters.
Dalman, Gustaf. 1905 (1982 (2)). Grammatik des judisch-palastinischen Aramaisch. Darmstadt: Wissen-schaftliche Buchgesellschaft.
Degen, Rainer. 1969. Altaramaische Grammatik der Inschriften des 10.-8. Jh. v. Chr. Stuttgart: Steiner. Eades, Domenyk. 2008. The Arabic Dialect of a Sawawi Community of Northern Oman. In Arabic Dialectology: In Honour of Clive Holes on the Occasion of His Sixtieth Birthday, ed. E. Al-Wer and R. de Jong. Pp. 77-98. Leiden: Brill.
Eisele, John. 1999. Arabic Verbs in Time: Tense and Aspect in Cairene Arabic. Wiesbaden: Harras-sowitz.
Fischer, Wolfdietrich. 1961. Die Sprache der arabischen Sprachinsel in Uzbekistan. Der Islam 36: 232-63.
Garcin, Jean-Claude. 1976. Un centre musulman de la Haute-Egypte medievale: Qus. Cairo: Institut Francais d'Archeologie Orientale.
Germanos, Marie. 2009. From Complementizer to Discourse Marker: The Functions of ?enno in Lebanese Arabic. In Information Structure in Spoken Arabic, ed. J. Owens and A. Elgibali. Pp. 145-64. London: Routledge.
Hasselbach, Rebecca. 2006. The Ventive/Energic in Semitic: A Morphological Study. Zeitschrift der deutschen morgenlandischen Gesellschaft 156: 309-28.
Hetzron, Robert. 1969. The Third Person Singular Pronoun Suffixes in Proto-Semitic. Orientalia Sue-cana 8: 101-27.
__. 1977. The Gunnan-Gurage Languages. Naples: Istituto Orientale di Napoli.
Holes, Clive. 1987. Language Variation and Change in a Modernising Arab State. London: Kegan Paul International.
__. 2011. A Participial Construction of Eastern Arabic: An Ancient Prediasporic Feature? Jerusalem Studies in Arabic and Islam 37: 75-98.
Ibn Hisham, Jamal al-Din. 1979. Mughni l-labib 'an kutub al-a'arib, Ed. Mazin Mubarak and Muhammad Hamd Allah. Cairo: Dar al-Fikr.
Ibn Mujahid, AbU Bala. 1972. Kitab al-Saffa fi l-qira'at. Ed. Shawqi Dayf. Cairo: Dar al-Ma'arif.
Ingham, Bruce. 1994. Najdi Arabic, Central Arabian. Amsterdam: Benjamins.
__. 2006. Afghanistan Arabic. Encyclopedia of Arabic Language and Linguistics, ed. Kees Versteegh, 1: 28-35. Leiden: Brill.
Jastrow, Otto. 1997. The Neo-Aramaic Languages. In The Semitic Languages, ed. Robert Hetzron. Pp. 334-77. London: Kegan Paul.
Kais, Dukes. 2011. The Quranic Arabic Corpus, version.2. http://corpus.quran.com/.
Landberg, Carlo, Comte de. 1905. Etudes sur les dialectes de l'Arabie meridionale, vol. 2,1. Leiden: E. J. Brill.
__. 1909. Etudes sur les dialectes de l'Arabie meridionale, vol. 2,2. Leiden: E. J. Brill.
Macuch, Rudolph. 1982. Grammatik des samaritanischen Aramaisch. Berlin: De Gruyter.
Meyer, Rudolph. 1972. Hebraische Grammatik, vol. 3. Berlin: Walter de Gruyter.
Moscati, Sabatino, et al. 19802. An Introduction to the Comparative Grammar of the Semitic Languages. Wiesbaden: Harrassowitz.
Nikolaeva, Irina. 2007. Constructional Economy and Non-Finite Independent Clauses. In Finiteness, ed. Irina Nikolaeva. Pp. 138-82. Oxford: Oxford Univ. Press.
Noldeke, Theodor. 1893. Bemerkungen zu den aramaischen Inschriften von Sendschirli. Zeitschrift der deutschen morgenlandischen Gesellschaft 47: 96-105.
Owens, Jonathan. 1993. A Reference Grammar of Nigerian Arabic. Wiesbaden: Harrassowitz.
__. 1998. Neighborhood and Ancestry: Variation in the Spoken Arabic of Maiduguri, Nigeria. Amsterdam: Benjamins.
__. 1999. Uniformity and Discontinuity: Toward a Characterization of Speech Communities. Linguistics 37: 663-98.
2003. Arabic Dialect History and Historical Linguistic Mythology. JAOS 123: 715-40.
__. 2009 (2). A Linguistic History of Arabic. Oxford: Oxford Univ. Press.
Rabin, Chaim. 1951. Ancient West Arabian. London: Taylor's Foreign Press.
Reichmuth, Stefan. 1983. Der arabische Dialekt der Sukriyya im Ostsudan. Hildesheim: Georg Olms.
Reinhardt, Carl. 1972 (1890. Ein arabischer Dialekt gesprochen in 'Oman und Zanzibar. ... Amsterdam: Philo Press.
Retso, Jan. 1988. Pronominal Suffixes with -n(n)- in Arabic Dialects and Other Semitic Languages. Zeitschrift fur arabische Linguistik 18: 77-94.
__. 1995. Pronominal State in Colloquial Arabic: A Diachronic Attempt. In Dialectologia arabica: A Collection of Articles in Honour of the Sixtieth Birthday of Professor Heikki Palva, ed. Tapani Harviainen et al. Pp. 183-92. Helsinki: Finnish Oriental Society.
Rosenthal, Franz. 1961. A Grammar of Biblical Aramaic. Wiesbaden: Harrassowitz.
Rossler, Otto. 1950. Verbalbau und Verbalflexion in den semito-hamitischen Sprachen. Zeitschrift der deutschen morgenlandischen Gesellschaft 100: 461-514.
Ibn al-Sarraj, Muhammad. 1985. Kitab al-Usul fi l-nahw. Ed. Abd al-Husayn al-Fatli. Beirut: Mu'assasat al-Risala.
Seeger, Ulrich. 2002. Zwei Texte im Dialekt der Araber von Chorasan. In "Sprich doch mit deinen Knechten aramaisch. wir verstehen es!": 60 Beitrage zur Semitistik. Festschrift fur Otto Jastrow zum 60. Geburtstag, ed. W. Arnold and H. Bobzin. Pp. 629-46. Wiesbaden: Harrassowitz.
Sibawayhi, 'Amr b. 'Uthman. 1970 (1898-99 (1)). Al-Kitab. 2 vols. Ed. H. Derenbourg. Hildesheim: G. Olms.
Sivan, Daniel. 1997. A Grammar of the Ugaritic Language. Leiden: Brill.
Soden, Wolfram von. 1952. Grundriss der akkadischen Grammatik. Rome: Pontificum Institutum Biblicum.
Testen, David. 1993. On the Development of the Energic Suffixes. In Perspectives on Arabic Linguistics V, ed. Mushira Eid and Clive Holes. Pp. 293-312. Amsterdam: Benjamins.
__. 1998. Parallels in Semitic Linguistics: The Development of Arabic -la and Related Semitic Particles. Leiden: Brill.
Tropper, Josef. 2000. Ugaritische Grammatik. Munster: Ugarit-Verlag.
Versteegh, Kees. 1997. The Arabic Language. Edinburgh: Edinburgh Univ. Press.
Wetzstein, I. G. 1868. Sprachliches aus den Zeltlagern der syrischen Wuste. Zeitschrift der deutschen morgenlandischen Gesellschaft 22: 69-194.
Williams, Ronald. 1972. Energic Verbal Forms in Hebrew. In Studies on the Ancient Palestinian World: Presented to Professor F. V. Winnett on the Occasion of His Retirement 1 July 1971, ed. J. W. Wevers and D. B. Redford. Pp. 75-85. Toronto: Univ. of Toronto Press.
Wilmsen, David. 2010. Dialects of Written Arabic: Syntactic Differences in the Treatment of Object Pronouns in Egyptian and Levantine Newspapers. Arabica 57: 99-128.
Wright, William. 1986 (1896-981). A Grammar of the Arabic Language, 3rd ed. 2 vols. in one. Cambridge: Cambridge Univ. Press.
Youssef, Zafer. 1990. Das Partizip im Arabischen: Die Auffasungen der arabischen Grammatiker und der Sprachgebrauch in klassisch-arabischen Texten. Ph.D. diss., Univ. of Erlangen.
Zaborski, Andrzej. 1996. On the Origin of Subjunctive and Energicus in Semitic. Incontri Linguistici 19: 69-76.
Zimmermann, Gent. 2009. Uzbekistan Arabic. Encyclopedia of Arabic Language and Linguistics, ed. Kees Versteegh, 4: 612-23. Leiden: Brill.
Zwettler, Michael. 1978. The Oral Tradition of Classical Arabic Poetry. Columbus: Ohio State Univ. Press.
Author's note: I would like to thank Clive Holes for comments on and discussion of various points in this article, and the two anonymous readers for their helpful and critical comments. Nadine Hamdan provided invaluable help in organizing the Qur'anic databases. Symbols used herein are standard grammatical abbreviations, e.g., N = noun, SG = singular, PL = plural, M = masculine, F = feminine, NOM = nominative, ACC = accusative, GEN = genitive, OBJ = object, NP = noun phrase, PP = prepositional phrase, etc.; in addition, EM = emphatic morpheme, T = energic suffix (al-nun al-thaqila), -n = intrusive -n, V = verb (unless in phonological formulae, where v = vowel), AP = active participle, CA = Classical Arabic. (Presumably) spoken dialect has been transcribed using a variant of the international phonetic alphabet (IPA) for consonants (dots under emphatic consonants instead of the IPA raised 'ayn) and two short vowels for a lengthened vowel (e.g., uu = [bar.u]).
(1.) Cf. Landberg's term, "Min paragogique" (1909: 732).
(2.) One of the readers observed that Barth's theory "... was conceived before the inclusion of Akkadian and Ethiopic ... in the reconstruction of [the] Proto-Semitic [-n]" and therefore can no longer be considered of equal rank with the inherited verbal suffix analysis. While this paper filters Akkadian out of the main analysis (see 5.5, below), the question of the origin of the intrusive -n as an originally independent element vs. original verbal suffix remains a question implicated for all of West Semitic (and ultimately all of Semitic). I do not think that the solution proposed here cannot be interpreted to have occurred at the proto-Semitic, rather than proto-West Semitic stage.
(3.) Carl Reinhardt's phonetic transcription is inconsistent. For the sound MPL suffix, for instance, he writes -yn, obviously transliterating rather than transcribing.
(4.) Rebecca Hasselbach's deriving the Bahraini geminate forms from analogical extension of the ISG object suffix (2006: 324) appears unnecessarily complicated; see (42) below.
(5.) Speakers of Afghanistani Arabic migrated from Uzbekistan in the nineteenth century.
(6.) In the third person the original AP forms marked by the number and gender suffixes are used.
(7.) Note, with the meaning "I have written."
(8.) Private communication, February 2011. See also Eades 2008: 89.
(9.) Hence in the ISG -ni, as well as -i.
(10.) Landberg's transcription has been adjusted to conform with the transcription of dialectal speech used in this article.
(11.) In the Arabic grammatical tradition, the final -n of the indicative (the inflection of indicative in these persons) is said to be dropped before the energic forms. William Wright (1986, 1:61) formulates the morphology differently, saying the energic is added to the jussive stem, which in the 2FSG and MPL at least is a viable alternative. It is not added to the jussive of weak final verbs, however, which maintain their long vowel before the suffix, e.g., armiy-anna "I shall throw" (Sibawayhi 1970.2: 161). There are interesting issues of historical linguistic interpretation behind both formulations.
(12.) Sibawayhi notes that in this case the light -n is not like the tanwin, which it behaves like in pausal position. The definite article is realized as al- post-pausally, hut as l- in context (way1). In the Arabic grammatical tradition, it is known inter alia as lam al-ta'rif "the definite l."
(13.) No attempt will be made to distinguish early forms of Arabic--pre-Classical, Classical, Qur'anic Arabic, and the like--which I believe is a precarious exercise (see Noldeke 1893, Fischer 1961; see also Owens 2009: 39-40 for methodological problems in defining early varieties). In my view, so long as a sociolinguistic history of early Arabic is lacking, issues of early varieties cannot be defined with complete adequacy.
(14.) http://corpus.quran.com/ (directed by Kais Dukes and hosted at Leeds University).
(15.) Viz., yakuna, noted in Sibawayhi as yakun-an (2: 152).
(16.) Given the discussion in n. 19, below, a more appropriate translation might be "and do you really think that his effort will remove that which enrages."
(17.) Similarly in poetry. Diwan 'Antara turned up nine total tokens of the energic -n, all of them marked with la- and none with the jawed, al-yasam actually "answering" an overt oath. My thanks to Smaranda Grigore for these statistics.
(18.) The verb is always in the 2MSG here.
(19.) A full-scale account of the semantics and pragmatics of the energic is outside the scope of this article. It can be noted, however, that on the basis of this characterization further interpretations relating to its meaning can be developed. For instance, looking at (16), the use of the yes-no question marker hal and the energic on a second person verb is pragmatically contradictory, as it combines a question (hal) with an assertive morpheme, -anna. The speaker is at one and the same time in good faith asking (with hal) for information from the addressee and asserting (with -anna) knowledge of the outcome. Note that the speaker is not asking whether the addressee is certain about what he is saying (which would be something like hal muta'akkidan taqulu dhalika "are you sure you are saying that = make sure that you know what you are saying is what you are saying"). Rather, -anna represents the speaker's (or third person subject's) own evaluation of the future event expressed in the predicate. On the basis of this pragmatic contradiction it would appear to be fruitful to derive the implied meaning noted in (16), above. Given that the sentence is grammatical, the inference is invited that the question the speaker is "asking" is not a question at all, which is a first step towards arriving at the illocutionary force of the configuration. The inference that it is a command rather than an assertion needs to be explicated.
(20.) From a brief comparative perspective, it can be noted that in a number of respects the energic is akin to what Irina Nikolaeva (2007: 141) terms hortatives. Hortatives generally occur in main clauses or, if in dependent contexts, then as clausal complements of verbs of speaking or epistemic verbs such as qala 'say' and alima 'know'; they tend to have reduced morphological contrasts, and they express exhortations or wishes. All three criteria are relevant to interpreting the energic.
(21.) Example (28) is a variant reading of innahum (Sa'id ibn Jubayr) (see Ibn Hisham, 307, for a discussion of acceptability of anna + la). In the later Arabic tradition, e.g., the summary of Ibn Hisham (300-11), the "lam of the oath" is distinguished from the "lam of the topic" (lam al-ibtida), which is illustrated in (15). Both have the function of emphasis, however, and both place la- on the predicate. Syntactically the Arabic tradition would be prone to consider the two lams distinctive, simply because of the different syntactic environments automatically generated by the nominal vs. verbal sentence--the former, Topic + (la-) Comment, the latter (la+verb-anna- Agent. The assumption is made here that speakers would have identified the two, both formally (both are la-) and semantically, even if they might have distinguished them syntactically, which is an open question.
In his exposition, Ibn Hisham recognizes four lams, homophonously la-, that do not govern and that occur on a major predicative constituent (verb, topic, subject), i.e., they contribute to the illocutionary force of the clause they occur in. These are the lam al-ibtide and lam (jawab) al-qasam, as noted, as well as the "extra lam" (al-lam al-zeida) and the lam that occurs on conditional particles. That the status of the four categories as distinctive categories is not always clear is witnessed by the fact that Ibn Hisham often notes that in various citations different grammarians differ as to the correct categorization of a given la-.
(22.) I.e., the morphemic material -anna is necessarily already in place before the analogical extension described in Figure 1 and (32) takes place. A reader suggests as a possible source constructions of the type yara anna 'see that' (> yara-nna). As will be specified in Figure 2 below, anna here is seen as providing "analogical help," but not as providing the morphemic material itself. Still, the existence of yara anna in discourse could further have expedited the formation of the analogy described in (32). V + anna is indeed quite common in the Qur'an (e.g., 2:26, 72, 106, 107, 194, 209, 223, 231, 233, 244, etc.), so this source deserves closer textual study as well.
(23.) The 3MSG is shown by labialization or palatalization in the non-nna- forms. Robert Hetzron (1969: 108) reconstructs the form as *-nn-uu.
(24.) Ronald Williams (1972: 78) has a rather different interpretation of Aramaic (and Hebrew) morphological structure, identifying all -n's after 2FSG and 2/3PL forms as tokens of the energic -n, as in the -n of taciabbal-uu-n "you (PL) shall receive" (see discussion below).
(25.) Williams (1972) points out the general lack of strict correlation between the presence or absence of -n and a grammatical category in Northwest Semitic languages, such as indicative vs. jussive. In Biblical Aramaic, for instance, ybaha-luu-k "they disturb you" and ybahlo-nna-ni "they disturb me," with and without -n, have no obvious difference in meaning.
(26.) Hetzron, and following him, Retso (1988: 87), has a different interpretation, whereby he identifies two [n + object] suffix morphemes in Hebrew, one related to the energic which occurs in all persons of the prefix conjugation and is not restricted phonologically, and one with a different origin determined phonologically in the context C + u/a + -nn + object suffix. In the 3SG the forms are contrastive, -nna/ennuu for what Hetzron (1969: 101) calls "-n suffixes," vs. n-hu/n-ha for the historical energic.
As Hasselbach (2006: 313) notes, postulating two different historical origins for -n's that occur in the identical context of V + n + pronoun object is to draw a perhaps overly fine diachronic distinction. Hetzron (1969: 112) identities the -n of the Gurage group (see (30), above) with the -n suffix of Hebrew (i.e., not with the historical energic -n). Whatever the merits are for making this association, Hetzron did not show comparatively with which of his proto-Semitic -n's the Aramaic and Arabic -n's are to be related. Ultimately, Hetzron projects an interesting synchronic contrast in Biblical Hebrew onto the entire fabric of proto-Semitic. While such inferences can hardly be ruled out in principle, they require systematic integration into an explanation for all languages in which a potential reflex of the form occurs.
(27.) The hedge in this formulation is the word "grammaticalized." The status of -n + noun in other West Semitic languages is discussed in the appendix.
(28.) Criteria for answering whether iyyd is a proper suffix or a clitic can be complex. Two points can be noted here based on Kuwaiti Arabic (thanks to information from Talal al-Jassar). First, -iyyti + pronoun forms a phonological word with the verb it is suffixed to, in that the last syllable of iyyti--carries the main lexical stress (following the universal Arabic stress rule of stressing the first heavy syllable from end). Second, iyya itself (no longer) occurs self-standing, at least in Kuwaiti Arabic, but only as an affix.
(29.) In this context it is appropriate to note that probably because of the common a element, it is not uncommon to see the energic posited as deriving from the Arabic subjunctive (e.g., Zaborski 1996: 70). This is not a position adopted here. The formal interpretation of -V n is the topic of this section. Semantically it was shown (see 2.2, 2.3, above) that the Arabic energic has a highly specific meaning defined inter alia by co-occurrence with a limited set of illocutionary morphemes, none of which is inherently associated with the subjunctive. Andrzej Zaborski makes no mention at all of the -n + pronoun object construction as a possible source of the energic.
(30.) Also cf. CA, stems II-IV with u vs. all others with a.
(31.) Wetzstein (1868) gives the vowel quality as a (see (4), above), and Ingham reports an a in Afghanistani Arabic.
(32.) The assumption that proto-Arabic had, for instance, a as preformative vowel, as in Sabatino Moscati et al. (1980: 141) and Kees Versteegh (1997: 99), is based on stipulations and does not explain on a comparative basis the widespread variation of the two vowels throughout grammatical morphemes. Both a and i occur in all regions of the Arabic-speaking world, with perhaps a greater propensity for a to occur in southern areas (Sudan, Yemen) and i in northern ones.
(33.) There are other -n's in Arabic that have the same allomorphy as (42). The FPL in those dialects that maintain it always have -a in-C ~ -a/inn-V, and the plural -iin in AP predicates often geminates to -inn-V. Various -n-final function words (e.g., min 'from') have a similar behavior.
(34.) It is striking that almost all contemporary observers who have had the benefit of tape recorders agree on the allomorphy in (42).
(35.) Why Biblical Aramaic should have a vowel before -n (inna-n) is a separate issue, for which a number of plausible parallels could be cited but are outside the scope of this article.
(36.) Thus, from the West Semitic perspective, the basic conditioning factor is the following segment rather than the preceding one, as some have argued for in Akkadian (Testen 1993: 305).
(37.) I owe the summary of the development of the eastern forms entirely to Clive Holes (private communication. February 2011).
(38.) This is a singular development. In other paradigms of Bagirmi Arabic SG/PL and MIF are differentiated consistently with morphology (see (7), above).
(39.) Linguistic observations relevant to this point are made below in 5.4.
(40.) Maltese appears to be the only "dialect" where its distribution is severely limited.
(41.) Its function as a predicate also broadly tallies with its usage in Classical Arabic (see Youssef 1990).
(42.) Variational data from Nigerian Arabic in Maiduguri is also interesting in this respect (Owens 1998: 289; Owens 1999).
(43.) In three corpora of spoken Arabic for which tagged data is available to me, two from Emirati Arabic and one from Nigerian Arabic, the AP as predicate has a frequency of about five percent of that of finite verbs (perfect and imperfect combined).
(44.) The two examples cited by al-Astarabadhi are not cited by Sibawayhi in his discussion of the nun, nor in book one in his chapter on poetic license (1970, 1: 7-10). The citation from al-AstarAbadhi, a grammarian who lived some five centuries after Sibawayhi, is to be interpreted as applying to the early Islamic era.
(45.) There are a number of descriptive and theoretical issues involved here. The basic issue is well defined, however. If the above reconstruction of the CA energic in 2.3 and the primacy of -n + pronoun object for PWS, as argued for in section 4, are correct, then the ventive as a verbal suffix (V + -n) in Akkadian requires independent reconstruction as well. The existence of the energic function in CA cannot be projected back into a PWS V+ -n construction. This is an issue I leave at this point to specialists in East Semitic. More generally, as has already been noted, stages 2 and 3 in Figure 2 may turn out to be proto-Semitic rather than specifically proto-West Semitic stages.
(46.) From the opposite direction, a similar problem arises with Hetzron's (1969: 110) assumption that the -nn object marker on the past (suffix) conjugation in Gurage is an independent development within proto-Ethiopic, for he did not account for the fact that the marker is widespread on the perfect verb in different varieties of Middle Aramaic. One can equally question Hetzron's assumption (1969: 124) that the intrusive -n on the AP in dialectal Arabic (he cites Omani) is an independent development. The only reason he adduces for this is that it occurs on the AP, not on the verb. However, the -n on the AP is already attested in Middle Aramaic and sound arguments for analogical extension to the AP can be given (see 5.4, above). Hasselbach as well (2006: 324) appears to countenance parallel independent development in the case of the geminate -nn; in her interpretation, in Omani and Bahraini Arabic (no doubt a single development in this case) and in Biblical Aramaic the gemination derives from -n + ni (1SG), in both cases that gemination spreading analogically to other persons. For West Semitic, a different development of the geminate form is argued here (section 5.2, above).
UNIVERSITY OF BAYREUTH
|Printer friendly Cite/link Email Feedback|
|Publication:||The Journal of the American Oriental Society|
|Date:||Apr 1, 2013|
|Previous Article:||Foreigners in the ancient Near East.|
|Next Article:||The fructification of the tale of a tree: the Parijataharana in the Harivamsa and its appendices.|