tionalists?". Zeitschrift der deutschen morgenlandischen Gesellschaft 134. 226-236. ... Kofler, Hans. 1933. "Handbuch des islamischen Staats- und Verwaltungs- .... arabische Grammatik sich als eine technische Disziplin herausbildete, hatte ...
THE NOTION OF 'UNDERLYING LEVELS' IN THE ARABIC GRAMMATICAL TRADITION* KEES VERSTEEGH University of Nijmegen
0.
Introduction
In his analysis of what he calls 'Cartesian linguistics' Chomsky (1966, and cf. 1965:198, n.12) points at a distinction between 'deep' and 'surface' structure in the Grammaire generate et raisonnee of 1660. In his view, this dichotomy correlates with that of mind and body in Descartes' philosophy, deep structure being the representation of a thought process, and surface structure being the representation of a speech utterance. By reconstructing the 'deep structure', Chomsky says, the Port Royal grammarians find out what the meaning of the utterance is, which may or may not correlate in a one-toone relationship with physical speech. Chomsky connects this operation with two aspects of Cartesian philosophy. In the first place, the relationship between language and thought, and in the second place, the underlying universality of thinking as against the contingency of language. The aim of the reconstruction of the underlying level is to demonstrate the homogeneity of mental categories and processes. These processes are translated by 'transformations' into a surface structure, and since the transformations differ from language to language, the physical utterances in these languages vary, too. The mental representations may be mapped onto the phonetic utterances, but since the linguist-philosopher is not overly interested in the exact nature of the correlation between the two levels, he has no need to study this relationship in detail.
* A preliminary version of this paper was presented at the 6th International Conference on the History of the Language Sciences (ICHoLS VI), held at Georgetown University, August 1993.I wish to thank especially one of the anonymous readers for his/her comments.
HistoriographiaLinguistica21:3 (1994), 271-296. DOI 10.1075/hl.21.3.02ver ISSN 0302-5160 / E-ISSN 1569-9781 © John Benjamins Publishing Company
272
KEES VERSTEEGH
It should be added that this reconstruction of Cartesian thought is not incontroversial. In many of the reviews of Chomsky's book it has been pointed out that his view of Cartesian linguistics and Port Royal grammar may have been coloured by his wish to demonstrate that there had been precursors for his own approach to linguistics other than his immediate post-Bloomfieldian predecessors. For my purpose it does not matter whether or not his analysis was historically or, for that matter, philologically, sound. Even if the 'Cartesian linguistics' as described in his book is just an artefact of his own views of the historical development of the discipline, it still stands as an interesting and relevant conception of the dichotomy between actual speech and underlying level, which may be used for comparative reasons. In the Arabic world a similar approach to language and thought was once current during a brief period in which Greek logic and philosophy were imported by a group of translators and philosophers. They wished to show, like the scholars involved in the Port Royal project, that mental processes are universal, and that the utterances are arbitrary and uninteresting from the point of view of the philosopher. This led to a series of famous and heated debates between grammarians and philosophers, each wishing to claim the domain of the ma'ânî, the 'meanings', for themselves. To us these debates are interesting, because they give us an insight into the concept of 'meaning' as it was used in the Arabo-Islamic world: for the philosophers the ma 'dni denote the universal significata of contingent speech, whereas for the grammarians ma 'dni are the essentially linguistic correlate of phonetic alfâz. In this paper I am concerned with the analysis of linguistic thoughts. As a result, I shall concentrate on the latter meaning of ma 'ânî, which is intimately connected to the concept of underlying level.1 My aim is threefold. In the first place I wish to show the connection between the disciplines of exegesis and grammar with respect to the interpretation of speech. In the second place, I briefly compare the Arabic approach with that of the Greek grammarians. In the third place, I shall make an attempt to look at the Arabic concept of underlying structure from the point of view of modern linguistics. 1.
From exegesis to grammar
The mainstream of Arabic linguistics is characterized by one feature which sets it apart sharply from the agenda of the philosophers in the Islamic period: they recognized only one language, which served as the initial and the ultimate model for their views on language and linguistic analysis. In the
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
273
Arabic tradition, the standard attitude towards other languages than Arabic -— whose existence as such was of course known: the first grammarian to describe the entire language, Sîbawayhi (d. 177/793), was himself a Persian, who had learnt Arabic as a second language — was one of neglect and disdain. Almost all grammarians reasoned that either the other languages had the same structure as Arabic, in which case it would be useless to study them, or they were different, in which case by definition the structure of Arabic was superior since it was the language that God had selected for his last revelation. In either case, the study of other languages was irrelevant, and it was left to the speakers of these other languages to learn the structure of Arabic and, if they wished, apply their knowledge of the Arabic grammatical model to their own language. Actually, this is what happened in the case of the treatment of Syriac, Hebrew, Coptic and even Turkic.2 The Arab grammarians themselves were concerned only with the Arabic language; consequently, their agenda was forced upon them by constraints that differed considerably from those in force in the universalist tradition mentioned above. In order to find out what determined their approach to language we have to go back to the earliest period of Islam, approximately the first two centuries of the Higra (i.e., 7th/8th centuries A.D.). During this period the main concern for the Muslims was the text of the Holy Book, which constituted their guide not only in the field of religion, but in the management of their everyday life as well. Obviously, the importance of understanding the text correctly became a predominant factor in their first scholarly activities. Almost right after the death of the Prophet, some of the Believers began to specialize in the interpretation of the Qur'ân, which as a revelatory text did not always lend itself to immediate understanding. Besides, the new inhabitants of the Arab empire spoke a kind of Arabic that was strikingly different from the language in the pre-Islamic period and there were many among the Believers who found it difficult to understand the text with its pre-Islamic flavour. In their hermeneutic activities the early commentators concentrated on understanding the text as such. To a lesser extent, they were also concerned with the conformity of the language of the Qur'ân to standard Arabic. We can, therefore, distinguish two trends in the interpretation: on the one hand, it paraphrases what the speaker — i.e., in the case of the Qur'ân, God — really meant. On the other hand, it attempts to translate the text into 'normal' language, i.e., the everyday language of educated speakers.
274
KEES VERSTEEGH
The first trend aimed at the interpretation of the ma 'nâ, the meaning of the Qur'ân, in the sense of the intention of the speaker. Paraphrases of words, phrases or whole verses are introduced by yurîdu "he intends to say", ay "i.e.", ya'ni "this means, he means". The constant re-phrasing of the text led to an awareness on the part of the commentators that there is a discrepancy between what one says and what one means. Very often the speaker articulates his thoughts in such a way that the meaning is hidden; in other words, there is an element of idmâr "hiding" in articulated speech. In the early commentaries on the Qur'ân idmdr is used almost exclusively as an explanation for semantic deletions in the surface structure (cf. Versteegh 1993:146151). As examples of the use of idmdr we may quote the following from Muqatil (d.150/767; Tafsîr I, 422.1 and II, 485.15): Q. 12/5 Id taqsus ru'ydka 'alâ ihwatika fa-yakîdû laka kaydan "don't tell your dream to your brothers, as a result they might ensnare you"; before the second clause the commentator inserts the exegetical remark fa-yahsidûka idmdr "for they might become jealous, deletion"; obviously the motive behind this insertion is semantic Q. 16/96 ma 'indakum "whatever there is with you"; the commentator adds min al-amwdl idmdr "of possessions, deletion"; here the addition is motivated by the wish to specify the general mâ ("whatever ... of possessions")
I shall return later in this paper to the importance of idmdr as a hermeneutic explanation; it remained in use in the earliest grammatical treatises, but acquired a new sense, that of syntactic deletion, that is to say, a deletion dictated by the formal relations between the elements of the sentence, rather than by semantic or contextual considerations. The second trend in the Qur'anic commentaries, the translation into everyday language, goes beyond the proper understanding of the text. In a Qur'anic verse such as Q. 25/59 tumma stawd 'aiâ l- 'arši r-Rahmdnu "then, the Merciful on the throne sat down" it is very easy to understand what the text means, but the word order is unusual. There can be only one motive for the commentator to paraphrase such a verse with tumma stawd r-Rahmânu 'aiâ l- 'arši "then, the Merciful sat down on the throne"
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
275
Apparently, the commentator wishes the language of the Qur'dn to conform to the language of the Arabs, the (idealized) speech of the ancient Bedouin. The situation of diglossia in the early Arab empire was such that any views about the language of the Arabs could only be an idealization, since very soon after the conquests nobody in the sedentary parts of the empire spoke the language anymore in the same way as it had been spoken in the pre-Islamic period. Still, commentators felt that there was such a thing as 'standard language' and in their paraphrases they adapted the language of the Qur'dn to this norm, apparently without worrying about the consequences this might have in the religious domain. In sum one may say that the main emphasis in the early commentaries was on the semantic reconstruction of the intention of the speaker with the help of a paraphrase of the text and in some cases the restitution of the effects of idmdr. As for the second trend in the exegesis of the Qur'ân, those places where the commentator 'corrects' the text to make it conform to the language of the Arabs are not very numerous, and even then the wish to avoid any possibility of misunderstanding, for instance in cases of irregular word order, may have been the prime motive for rephrasing the text. 2.
Underlying levels in early Arabic grammar
The emergence of grammar as a specialized discipline took place in the second half of the second century of the Higra (roughly the period between 770-820 A.D.). We know that among the first grammarians in the period before Sîbawayhi there were those who went even further than the commentators and explicitly attempted to correct the language of the Qur'ân according to the qiyâs al- 'Arabiyya, the rules of the Arabic language. By the end of the second century of the Higra this attitude had disappeared completely, largely because of the influence of Sibawayhi and al-Farra' (d.207/822): both grammarians accepted the codified text of the Qur'dn as the ultimate model for correct speech and rejected any proposal to change its wording (cf. Versteegh 1983). But even these grammarians had to deal with the fact that there was a difference between the accepted language of their Holy Book and the 'normal' language. The second trend in the interpretation of the Qur'dn thus led to a crucial shift in the development of linguistic thinking, the awareness that a thought may be articulated in different ways, which are ordered hierarchically. As the Qur'dn itself emphasizes, God's speech has been revealed in the
276
KEES VERSTEEGH
Arabic language, and consequently, it exhibits variation along the same patterns as the speech of human speakers. Within this variation there is a hierarchical ordering, some variants being perceived as less usual than others. This means that it is perfectly legitimate to compare the language of the Qur'dn with standard speech and point out the deviations that fall within this variation. Consequently, the varying word order and similar syntactic phenomena in the text of the Qur'ân are regarded as a derived way of speaking, i.e., derived from the 'normal' word order by various articulatory processes, e.g., the transposition of words (in the commentaries the technical term for this procedure is taqdîm or muqaddam wa-mu'ahhar), or the deletion of units (idmdr). The acceptance of the codified text of the Qur'ân by the community as well as the scholars not only prevented the grammarians from interfering with the text, it also led to a shift of interest on their part. Instead of concentrating on the text of the Holy Book, they started to analyze the structure of the language. For grammarians such as Sîbawayhi the language of the Qur'ân remained the ultimate source for correct language, but since it was no longer possible to make changes in the text or to compare variant readings, the grammarians became much more involved in the analysis of the canonical structure of language. They no longer bothered with the lexical meanings of items in the language, either, since that task had been taken over by specialists in Arabic lexicography ('ilm al-luga). What the grammarians took over from the commentators was the idea that underlying the level of actual speech there was a level of analysis. The commentators needed this level in order to interpret the intention of the speaker. The grammarians, however, were concerned with the structure of the sentence. They compared canonical and derived sentences in connection with the theory of governance within the sentence. We do not know with any certainty when exactly the concept of 'amal "governance" was introduced in the study of language. There is no trace of it in the commentaries, but in Sîbawayhi's Kitdb it is already fully developed (cf. also Talmon 1993). The consequences of this theory for the analysis of the structure of actual utterances cannot be overestimated. If a change in a declensional ending must be related to a governor, as the theory dictates, the grammarian must be able to point out a governor for each declensional ending. Since in many instances the actual sentence does not contain an overt governor, by necessity the analysis has to take recourse to an underlying level.
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
277
Take, for instance, the sentence zaydan darabtu "Zayd I hit": the commentators would find no difficulty with the meaning of this sentence, but if for whatever reason they felt the need to paraphrase it, they did so with the help of the hermeneutic device of the taqdim, stating that what was meant was darabtu zaydan. The grammarians, however, were confronted with a problem: darabtu is governor, but it is preceded by the word governed by it, something which their theory precludes. Therefore, they took over the analysis of the commentators by applying the device of taqdîm and stated that underlyingly there was another sentence, in which darabtu precedes as governor the object/accusative zaydan. If we take a different sentence, zaydan darabtuhu "Zayd, I hit him": here the device of taqdîm does not work, since the result would be the ungrammatical sentence *darabtuhu zaydan. Consequently, the accusative must have a different governor, which is absent in the actual utterance, but must be reconstructed in the underlying sentence: *darabtu zaydan darabtuhu. This is, of course, an awkward sentence, but it does show clearly the relations between governors and governed words, and thus satisfies the theoretical needs of the grammarian. The notion of 'deletion' is indicated in Sibawayhi's Kitâb by the term idmdr, literally "hiding", which we have noted earlier with regard to the Qur'anic commentaries. This notion of idmdr is analyzed extensively by Ayoub (1990). It indicates on the one hand the anaphorical personal pronoun, or more exactly the deleted noun to which the pronoun refers: al-asma' allati hiya 'alama li-l-mudmar "the nouns which are a sign to the hidden [noun]" (Kitdb I, 257.14),3 and on the other "des elements sans valeur phonique, identifiables en termes categoriels et ayant un effet sur la sequence phonique, cet effet etant un 'amaF (Ayoub 1990:4). The governing effect of the deleted element is mentioned by Sîbawayhi, for instance, in the case of the deleted verb which governs an accusative object (Kitdb I, 149fi'l mudmarlfi'l muzhaf). As a general term it may come to denote the concept of 'implicit speech'. The idmdr of the speaker has its counterpart in the grammarian's tamtil, the representation of the utterances in such a way that they become analyzable. One of the devices the grammarian has at his disposal is the restitution of the idmâr, e.g., tamtîl
zaydan marartu bihi ga 'altu zaydan 'ald tariqi marartu bihi "*I put Zayd on my way I passed him" "as for Zayd, I passed him"
278
KEES VERSTEEGH
Tamtil is more than just the restitution of deleted elements, however. It also involves the restoration of the canonical word order, e.g.: tamtil
zaydan darabtu darabtu zaydan "I hit Zayd"
In other instances the representation uses different methods, which may even include the omission of elements from the surface sentence, e.g.: tamtîl
ma sana 'ta wa-ahdka *mâ sana 'ta ahâka "*what did you do your brother?" "what did you do to your brother?"
In this example the particle wa- "and" in the surface sentence must be omitted on the underlying level, since it prevents the verb from exercizing its governance. The asterisk indicates the fact that the utterance in the representation is of a type that 'is not spoken' (lâ yutakallamu bihi), in other words, the elements contained in it are used solely for the purpose of representation and analysis, as a kind of metalanguage, just like abstract notations may be used to indicate the relations within the sentence. The attribute 'is not spoken' has nothing to do with the acceptability of the sentence, but simply indicates that this sentence does not purport to represent actual speech. It is important to note that in the examples given above the relation between the tamtil and the actual utterance is not one of paraphrase: the representation does not constitute a semantic reconstruction of the sentence. This becomes particularly clear when we look at the meaning (ma 'nâ) of the third example, which is not identical with its tamtil. Unlike the abstract representation the semantic paraphrase of the sentence is a complete, 'utterable' sentence: tamtil ma 'nd
mâ sana 'ta wa-ahâka *md sana 'ta ahâka ma sana 'ta ma 'a ahika "what did you do with your brother?"
Obviously, the term for 'deletion' in Sîbawayhi's Kitâb, idmâr, is connected with the idmâr of the commentaries, even though the purpose of the device differs. In the commentaries the main effect of restitution of the deleted element is the reconstruction of the intention of the speaker. In the
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
279
Kitâb, on the other hand, deletion is an essential factor in the analysis of the formal, syntactic relations between elements of the sentence, and the reconstruction of the deleted element is necessary in order to explain the effects of governance ('amal) in the sentence. We have seen above that the semantic interpretation of the sentence is not identical with the syntactic reconstruction. This is also illustrated by some instances in which according to Sîbawayhi the underlying meaning differs from the actual utterance, although the formal relations between the elements of the sentence are perfectly clear. In such instances Sîbawayhi attributes the discrepancy between meaning and utterance to the flexibility (ittisâ') of the Arabic language (cf. Versteegh 1990), which allows the speakers to use their own judgment in speaking. This may involve abbreviating speech and replacing elements by others. The difference with deletions is that the elements the grammarian adds in order to interpret the meaning of the sentence do not exercize any syntactic action in the surface sentence, whereas in deletions the deleted element in the underlying syntactic structure exercizes a syntactic action Carnal), which is manifest in the surface sentence and is, in fact, the main motive for positing a deleted governor. As an example we may quote the Qur'anic verse Q. 12/82 (STbawayhi, Kitab I, 108) isal al-qaryata "ask the village" This is interpreted as isal ahla l-qaryati "ask the people of the village" In this example the phrase "the village" has replaced "the people of the village", but the word "people" that was omitted in abbreviation by ittisd' does not operate on any of the words in the surface sentence. In itself the syntactic structure of the surface sentence is perfectly in order. In the writings of the Kufan grammarian al-Farra' (d.207/822), who is a near contemporary of STbawayhi, the term idmdr occurs very frequently (derivates from the root d-m-r altogether 139 times, the term idmdr itself 69 times, cf. Devenyi 1990:191-193), but it seems that he uses it much more in the old sense in which it was used in the commentaries, e.g., (Ma'âni I, 271 4): wa-dalika min kalâm al-'Arab an yudmirû man fî mubtada' al-kaldm fayaqûluna minnd yaqulu ddlika wa-minnd la yaquluhu wa-dalika anna man ba 'd li-md hiya minhu fa-li-ddlika addat 'an al-ma'nd al-matrilk "and this is one of
280
KEES VERSTEEGH the properties of the Arabic language, that they delete man 'who' at the beginning of the sentence, saying minnd yaqulu ddlika wa-minnd ld yaquluhu [i.e., minnd man yaqulu ddlika etc.] 'some of us say this, others don't' [i.e., "there are among us those who say this" etc.]; the reason is that [the preposition] min "of, from" indicates a part of the group to which it belongs and that is why it expresses the meaning that is left out".
Idmdr is the general term for any kind of deletion, of a word as in the example quoted, but also of a phonetic element as in (Ma 'tint I, 284.8) wa-kull mawdi' igtama 'a fihi td 'dni gdza fihi idmdr ihddhimâ "in every position in which there is a cluster of two r's, it is permitted to delete one of them", or of a clause as in (Ma'dni I, 273.13, 274.5) li-gazd' mudmar "for a deleted apodosis". The restitution of the deleted element is necessary in order to understand the meaning of the sentence, but it is not needed for the analysis of the formal relations within the sentence. In this respect al-Farra', unlike Sîbawayhi, represents the old tradition with its emphasis on the semantic reconstruction of the underlying intention of the speaker. We have seen above that in cases of purely semantic reconstruction, Sibawayhi does not use the term idmdr. 3.
The notion oftaqdîr
In Sîbawayhi's Kitâb the idmdr is the process by which the speaker leaves out elements in the sentence, which the grammarian then reconstructs through tamtîl. In later grammatical treatises, the terms idmdr and mudmar continue to be used for both the anaphorical (personal) pronouns and the process of deletion. But a new term is introduced to indicate the reconstruction of the underlying level by the grammarian, regardless of whether it differs from the actual utterance by deletion or through any other means (as in the case of mâ sana'ta wa-ahdka, mentioned above). This new term, taqdir, encompasses all instances of reconstruction of an underlying sentence, on the phonological, morphonological and syntactic level. If an element is said to be muqaddar, this means it is 'virtually there', but does not occur in the actual utterance. Although, lexically, taqdîr indicates the action of the grammarian who reconstructs the underlying level, it may also denote the underlying level itself. It is possible, for instance, to say that an utterance may be representedfî t-taqdîr in such-and-such a way, i.e., in the reconstruction by the grammarian, but also that its taqdîr is such-and-such, i.e., its underlying level. Since the term taqdir does occur in Sibawayhi, although in a different sense, the question arises as to what the terminological relation between taqdir
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
281
and tamtil is, and why the former completely replaced the latter in the analysis of the underlying level. Sîbawayhi uses taqdîr a total of 23 times (cf. Carter 1968:296-297). In only three instances it is used in a syntactic context, the other ones are phonological in nature. The syntactic examples are as follows: Kitâb I, 287.14: the structure of a-laysa hâdd zaydan muntaliqan "is this not Zayd, leaving?" is similar to daraba 'abdullâhi zaydan qd'iman '"Abdallah hit Zayd standing" in the taqdîr, but not in the ma'nd. Kitdb I, 300.14: inna-hu ddhibatun amatuka "your female slave is going away" is inna l-amra ddhibatun amatuka. Sibawayhi adds ka'annahu fi t-taqdir wa-in kâna lâ yutakallamu bihi "as if in the taqdir it is like this, even though it is not spoken this way". This construction involves the so-called damir aš-ša'n, which acts as a dummy pronoun at the beginning of the sentence (cf. Peled 1990). Kitdb I, 301.18: 'abdulldhi ni'ma r-ragulu "'Abdallah, what a man he is!" has the same taqdir, but not the same ma'nd as 'abdulldhi dahaba ahûhu '"Abdallah, his brother left".
All other examples of taqdîr are connected with the phonological structure of words, specifically of the glottal stop (hamza), whose phonological status is sometimes difficult to determine, since it may be part of the base structure, but also the result of a surface rule. Sîbawayhi establishes the status of the hamza by presenting the pattern of the word, for which he uses the term taqdir. The word gay'al, for instance (Kitdb II, 21.5), is said to have the taqdir gay'al, the word š a n û a has the taqdir sanû'a (Kitdb II, 70.19). In these and similar examples the term taqdir apparently indicates the structure of the word, and the letter 'ayn is used to indicate that the hamza has the status of a radical in the base structure. In one example (Kitdb II, 231.16) this becomes clear when it is said that the word (sâ'ani) suan has the taqdîrfulIan. In this case, the structure of the entire word is expressed with the help of an abstract notation with the lettersf- l,4 whereas in the other examples only the hamza is replaced by the 'ayn, without changing the rest of the word. What is the connection between these phonological examples and the three syntactic ones quoted above? In the last of the three syntactic examples above, the taqdir represents the structure of the sentence by using a notation in concrete words that indicates the functions in the sentence, in the same way as the phonological examples use concrete consonants to indicate the status of the hamza. In other words, the sentence, 'abdulldhi ni'ma r-ragulu, has the structure of mubtada — fi'l — fâ'il "topic - verb - agent", just like the sentence quoted as its equivalent in the taqdir. Sibawayhi adds that this is not the meaning of this sentence, but only its taqdir.
282
KEES VERSTEEGH
The original meaning of taqdir is the assignment of a qadr, a status. In theology, the usual meaning of taqdîr is "to determine", when used of God, and "to act consciously, with premeditation" or "to 'calculate' one's acts" (cf. Gimaret 1980:9-10, and n.3), but these meanings are not relevant here. In grammar, taqdîr functions as the verbal noun of a denominative verb qaddara "to assign qadr". Qadr is approximately the same as hukm, which indicates the judgment that is based on the status of an element. In Badr ad-Dîn Ibn Gama'a (d.733/1333) we find a similar meaning, when taqdir is used for assigning an allotment in the booty on the basis of a person's status (cf. Kofler 1933:384), and to cite a much earlier example, Ibn al-Muqaffa' (middle of the 2nd/8th century) uses taqdir in the sense of "appraising, appreciating", i.e., "assigning to someone his proper place and status" (Risâla 22). In linguistics, this is precisely what taqdir means in the early period: the assigning of a status. On the phonological level, this means the assigning of a status in terms of /-'-/, and on the syntactic level, it means the assigning of status in terms of functions within the sentence. There is no trace of any association with elision, deletion or virtuality. In later linguistic treatises, however, taqdir has replaced the notion of tamtil almost completely, and become the normal term for the underlying virtual level of analysis, sometimes in opposition to tahqiq "reality" in the sense of "actual utterance" (e.g., as-Sirbînî, Carter 1981:392.2). It may be added that taqdir does not only exist in linguistics: in the legal sciences it is used for the notion of 'legal fiction', i.e. "donner a l'inexistant (ma'dum) la qualification legale (hukm) de l'existant (maujud), ou a l'existant celle de l'inexistant" (Brunschvig 1970-72:44). In this context, too, it is the opposite of tahqiq. We do not know the exact development of the term taqdir, but so much is certain that in the work of Sîbawayhi's contemporaries and early successors among the Basran grammarians the term taqdir underwent a gradual change in meaning. Apparently, the term was not used in Kufa: al-Farra' uses taqdir only in Qur'ânic quotations in its non-linguistic sense of "decree by God" (cf. Devenyi 1990:194). In Abu 'Ubayda's (d.210/825) Magaz al-Qur'an the term is rather more common; it is used almost exclusively for morphological patterns (e.g., Magâz I, 170.4, 6; I, 202.2 etc.; cf. also the expression 'aid taqdirihâ, meaning "of this pattern", II, 153.6). In a few passages the meaning comes closer to that of a syntactic underlying pattern (Magaz I, 372.6: II, 196.6), but usually the underlying level, as far as deletion is concerned, is indicated with the term id mar, as in the Kitâb Sibawayhi.
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
283
In al-Mubarrad (d.285/898) taqdîr continues to be used in the same sense we found in Sîbawayhi, e.g., (Muqtadab I, 30.6) fa-sara taqdîruhu min al-fi'l laf'â 'u "expressed in the notation/- '-l its pattern becomes that oflaf'âV . 5 But it is also used in the sense of equivalent structure, e.g., the paraphrase of a masdar with an + verbal sentence {Muqtadab I, 14.1,2; I, 21.12) or the paraphrase of a participle with a relative sentence with alladî (Muqtadab I, 24.2).6 The essential point is that in al-Mubarrad, too, there is no trace of elision or virtuality in the meaning of taqdîr, for these notions he, like Sîbawayhi, uses the term idmdr. In both Sîbawayhi and al-Mubarrad the term taqdir is associated with an underlying level, but with a structural one, rather than one that is arrived at by restitution of deleted elements. There certainly is no relation with any underlying semantic interpretation, since according to Sibawayhi, taqdir is not connected with the ma'nd, and apparently for al-Mubarrad it is, neither. The most reasonable explanation for the fact that taqdir is not connected with meaning may be found in the distribution of the examples in the Kitâb. Since the majority of the occurrences concern phonological patterns, we may assume that originally, taqdir was connected with the phonological/morphological analysis of words, in all probability expressed in the/-'-l notation (cf. also note 3 of this paper). 4.
The underlying level in later grammar
In the preceding paragraph we have seen that in the later tradition there was a shift in the meaning of the term taqdir, which came to denote an underlying semantic level. In this section we shall take a look at the status of the underlying level in later theory. In general, Arab grammarians do not talk about this status, and they take the taqdir for granted as a device to clarify the structure of the surface structure. The only grammarian to discuss the status of the taqdir explicitly is Ibn Ginni (d.392/1002). He discusses the verbal form qâma "he stood up", whose phonological representation is *qawama. According to Ibn Ginni there are people who interpret the relationship between qdma and *qawama, its taqdir, in terms of a historical development (cf. the analysis by Guillaume 1981). But, as Ibn Ginni explains, such an interpretation is incorrect: it was never true that people used the forms that grammarians posit as underlying forms in order to explain the overall structure of the Arabic language. A verb such as ista'dra is different in surface form from istahraga, but since both forms function in the same way, as 10th measure to
284
KEES VERSTEEGH
a verbal stem, there must be an underlying structure *ista'wara. In other words, setting up an underlying form is necessary in order to achieve a coherent theory (cf. also Bohas 1981; and for an interpretation of the taqdîr in Ibn Hisam, see Gully 1991). Probably, IbnGinni's arguments against a 'diachronic' explanation were purely theoretical, since no one ever seriously attempted to give such a diachronic interpretation of the phonological representations. These had always received a 'synchronic' interpretation, although the theoretical basis for such an interpretation remained implicit. In the 3rd/9th century, however, the challenge from the imported Greek sciences of logic and philosophy became so pervasive that the grammarians were forced to explicitate their principles and methods, including the status of the taqdir. According to Guillaume (1981:233) this development started with Ibn as-Sarrag (d.316/928). The absence of any interest in 'diachronic' explanations is explained by the entire cultural and religious context, which argued against any development in the language. I have stated above that after the period of the conquests the Arabic language went through a series of drastic changes. But this statement was made from a typically Western historicizing perspective. The Arab grammarians formulated the development in an altogether different manner. According to them the new inhabitants of the Arab empire threatened to corrupt the language by making many mistakes while speaking Arabic. As a matter of fact, the emergence of grammar is often linked with this threat. Colloquial expressions of Arabic are, therefore, regarded by the grammarians as linguistic errors, that hold no interest for the specialists of language study, and in their theories there is no room for a vernacular as a linguistic variety sui generis. The grammarians were solely interested in the study of the standard language, i.e., the idealized high variety in a situation of diglossia, which supposedly remained unchanged throughout the centuries. Since the underlying level in phonology did not serve to explain the diachronic development of words, what then was the use of positing an underlying level? As Guillaume (1981) explains in his analysis of the status of the underlying level, its function was the simplification of linguistic theory by restricting the number of allowed structures. In syntactic matters, too, the positing of an underlying level achieved this aim. We have seen above that the sentence zaydan darabtuhu, which could have posed a problem in a unified description of Arabic sentence structure, was explained away by positing a deleted verb in initial position, that was responsible for the accusative in
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
285
zaydan. In Sibawayhi, this analysis was performed by the grammarian's tamtil. The operation of tamtil, which is Sibawayhi's parallel to the later taqdir, aims at preserving coherency in the theory. Through the reconstruction of constituent structure in the form of 'unspoken' sentences the grammarian can uniformize his description of the surface sentence and clarify the relations between its constituents. In the Kitâb, the notion of taqdîr seems to have had a different function, although we cannot be completely certain in view of the lack of sufficient examples. At this stage, taqdir simply represented the exact structural description of an actual string of sounds or words. After Sibawayhi it took some time to work out the terminology for underlying levels. The term idmdr remained in use as the usual term for 'elision' as a synonym for hadf (cf. also Carter 1991), e.g., in Ibn Ginnî's analysis of hadf (Hasd'is II, 360.2-3) in the bâb fi saga'at al-'arabiyya "chapter about the courage [!] of the Arabic language", which comprises hadf "elision", ziyâda "addition", taqdîm wa-ta'hir "transposition", al-haml 'aid l-ma'nd "constructio ad sensum", and tahrif "alteration". In this chapter mudmar seems to be a current synonym for mahduf(e.g., II, 375.5; II, 379.13 and passim on II, 380). Ibn as-Sarrag (Usûl II, 247ff.) distinguishes between three kinds of idmdr. zâhir la yahsunu idmâruhu "overt speech which may not be deleted", mudmar musta'mal izhdruhu "deleted speech which may also be used overtly", and mudmar matruk izhdruhu "deleted speech, which is never used overtly", of which only the third corresponds to the structural kind of deletion we have discussed above. The main condition he poses for all kinds of idmdr is that there must be in the linguistic or extralinguistic context (hâl al-musdhada) some evidence (dalîl) for the deleted element. The term tamtil is used only infrequently in later grammatical theory, and seems to have been subsumed under the term taqdir, which became the current term for positing an underlying level. In Ibn Ginnî taqdir is used either as an interpretation of the intention of the speaker, in which case it is related to or almost synonymous with ma'nâ (for instance, Hasd'is II, 362.11 where taqdir is even used in the case of ittisâ', so that it becomes identical with ma'nd), or as a reconstruction of the underlying structure in order to explain the surface structure. Still later taqdir meant simply 'virtuality', so that muqaddar became synonymous with what used to be called mudmar. The latter term tended to be reserved in later grammar for '(anaphorical) pronoun'. The structural meaning of taqdir in later grammatical works is very much akin to the meaning of tamtil in Sîbawayhi's Kitâb: it does not indicate the
286
KEES VERSTEEGH
relationship between the two levels in terms of a derivation (there are, for instance, no chains involved), but in terms of an interpretation of the surface structure. The crucial point is that the underlying structure is needed for the explanation of the syntactic relations in the surface sentence. This becomes very clear in the following examples discussed by Ibn Ginnî (Hasâ'is II, 384.13ff.): gi'tu râkiban tibtu bihi nafsan
"I came riding" "I was good to him in mind [i.e., I was well-disposed towards him]"
Superficially, these two sentences contain the same structure with a modifying accusative. But Ibn Ginnî observes one difference between them: it is possible to front the accusative in the first sentence, but not in the second: rdkiban gi 'tu *nafsan tibtu bihi This difference is explained by him through the reconstruction of their taqdîr (which he also calls asl,primary structure): whereas in the first sentence the accusative remains a circumstantial object, the second sentence undergoes a change in that the accusative becomes the agent; thus: gi'tu râkiban, not * gâ'a râkibi tâbat bihi nafsî
"my riding came" "my mind was good to him"
This explains the fact that the first accusative may be fronted, since it is not an agent in the taqdîr, whereas the second accusative cannot be fronted, since it is the agent in the taqdir and an agent can never be fronted before its verb. In other words, Ibn Ginni sets up different underlying levels in order to explain the difference in syntactic behaviour between the surface sentences. It would appear that by contrast, in modern generative linguistics the difference in syntactic behaviour serves as evidence for a different semantic interpretation necessitating the setting up of different abstract base structures. 5.
Conclusion
This last remark concerning modern generative linguistics touches upon a crucial problem in the interpretation of ancient linguistic theory, or, for that matter, any linguistic tradition that does not belong to the Western linguistic tradition. When analyzing the Arab grammarians' approach, one often tends
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
287
to compare their approach with a contemporary model. It is, however, not at all obvious what the value of such a comparison might be. There is a quite acute temptation for historiographers of linguistics to interpret the past in terms of the prevalent model in modern times. Thus, scholars have found precursors for generative and transformational linguistics in such diverse grammarians as Pâ.ini, Apollonius Dyscolus and Sîbawayhi. It is important to ask ourselves what purpose such a comparison might serve. If our aim is to legitimize a modern theory, then of course, we are not dealing with a historiographical analysis, but with gathering support for our own theory by aligning ourselves with a revered tradition. But if our aim is the interpretation of the past tradition, with a view to making it more easily accessible to the modern reader, reference to a contemporary model may provide us with the tools to bring out differences and resemblances, which might have remained hidden otherwise. For the Arabic tradition a formal comparison has been carried out by Owens (1988) with the express aim of bringing out the differences and resemblances in constraints that exist between the Arabic concept of deep structure/transformations and modern linguistic theory; a similar attempt was made by Gruntfest (1984), who, however, wants to show that the Arab grammarians were transformationalists avant la lettre. To begin with, there is one fundamental difference between the Arabic concept of the relationship between actual speech and underlying level, on the one hand, and the Western concept of surface and deep structure, on the other, in that the Arab grammarians did not formulate this relationship in terms of derivation chains or a process (a similar conclusion is reached by Gruntfest 1984:236). In general, we must say that the underlying structures are interpretive in semantic terms, rather than derivative.7 It seems that the Greek tradition, too, used interpretive or paraphrastic underlying structures, rather than derivative chains (cf. Householder 1981:4-17). Owens' conclusions (1988:196-198) regarding the differences in emphasis between the Arabic concept of underlying structure and the deep structure of generative linguistics may be summarized as follows. According to him there are four such differences. In the first place, generative grammar deals almost exclusively with textual context in conditioning deletion, whereas in Arabic grammar the pragmatic context may be just as important. In the second place, "ellipsis in Arabic theory is orientated towards reconstituting lost items whereas in transformational grammar it is orientated towards
288
KEES VERSTEEGH
decomposing complete ones" (p. 196). According to Owens taqdîr "always proceeds from a form which does not conform to the established grammatical rules and/or to an obvious semantic interpretation [...] to a more complete form which does" (p. 197). In the third place, in Arabic theory deletion is always linked to the speaker, without any formal rules of deletion specifying the conditions. In the fourth place, Arab grammarians used "recoverability of meaning as the limiting condition of ellipsis" (p. 198), a condition which is not shared by all versions of transformational grammar. Owens (1988:198) concludes: In the final analysis, these are differences of emphasis more than differences of substance because the basic motivations behind the use of deletions [are] the same: the desire to maintain an overall structural coherency in the grammar and to derive the correct meaning.
I do agree with Owens that this characterizes the Arabic use of deletion and underlying structure, but I am not so sure that it applies to the underlying motive for transformations in modern linguistic theory as well. It does seem to apply to Greek grammar, if we follow Blank's (1982) conclusions about the principles of Apollonius Dyscolus' syntactic theory. Blank states (p.52): Apollonius analyzed cases of irregularity (akatallêlîa) in order to demonstrate that they had originally followed the rule, but became corrupted in certain recognizable and corrigible ways, and in order to point out what the set of phenomena is, for the explanation of which rules must be set up.
The rules are there to explain the irregularities, which are shown to be deviations from the regular constructions. It is certainly relevant to point to an another observation made by Blank (1982:82) when he says that it was not the intention of the grammarian to correct the irregularities, quoting Apollonius as saying that the Homeric irregularities that have been transmitted (parádosis) must be retained. This reminds us of the attitude taken by the Arab grammarians vis-a-vis the Qur'anic text. In conclusion, we may say that in Arabic grammar the most salient feature of the theoretical concept of underlying structure or taqdîr in the later theory (Sîbawayhi's tamtîl) is the direction of the derivation: the transformations, if we wish to call them thus, are interpretive rather than generative, just like the transformations in most of Householder's examples from Apollonius Dyscolus. This was very much obvious in the example from Ibn Ginnî's Hasa'is we discussed above. If I am not mistaken, this difference is based on the basic difference in purpose: the Arab grammarians had no intention of
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
289
reconstructing underlying universal semantic interpretations, but they aimed at an explanation of the surface structure of just one language. Perhaps because of the exegetical background of the Arab linguistic tradition, or because of the experience they had had in debates with philosophical opponents whose claims went much further, they absolutely refrained from making any inferences about the mental processes involved and stayed within the realm of linguistic explanations. It is probably no coincidence that Greek grammarians, such as Apollonius Dyscolus, did exactly the same thing, since they, too, had been trained in the interpretation and edition of the Homeric poems, and they, too, had had their quarrels with philosophers, who claimed to know everything about language. It is true that Householder (1981) asserts that there are some indications of transformational thinking in Apollonius Dyscolus, but on the whole I think Sluiter (1990:46) is right when she says: Now, if we compare the Stoic theory with Apollonius' description of the structure of language, it is at once clear that, althought the former is clearly at the basis of the latter, the emphasis lies elsewhere. Both parties are interested in the relationship between expression and meaning, but Apollonius wants primarily to explain the correctness of the expression, whereas for the Stoics the theory of the lektón serves the twofold purpose of making a corrective addition to the ontology of the Peripatos and of supporting their logic.
This conclusion also points to the difference between the philosophical and the linguistic approach towards underlying levels.8 Both in the Greek and the Arabic tradition much remains obscure with regard to the theoretical presuppositions that guided their analysis, but were hardly ever made explicit. Obviously, there were many differences between the two traditions; for instance, in their approach to the semantic component of grammar and to the diachronic dimension of linguistic reconstruction. Greek grammatical theory was largely semantically oriented (cf. Sluiter 1990:2) and included diachrony and dialectal variation in the set of phenomena to be explained, whereas Arab grammarians were much more interested in the formal aspects of linguistic explanations, and remained within a strictly synchronic, invariable framework. But in other respects the two traditions were very much alike. The main conclusion I draw from this (very summary) comparison of the two traditions with respect to the status of the underlying level is that both groups of grammarians shared an interest in the linguistic dimensions of speech, with the almost complete exclusion of philosophical interpretations in terms of a dichotomy language/thought. When they posited
290
KEES VERSTEEGH
an underlying level, they did so purely in order to explain the actual speech phenomena they observed in their respective corpora of evidence. In this respect they were very much different from the universalist position referred to at the outset of this paper.
Author's address: Kees Versteegh Institute of Middle Eastern Studies University of Nijmegen Erasmusplein 1 NL-6500 HD NIJMEGEN The Netherlands NOTES 1) For a detailed analysis of this debate, which represented a crucial episode in the development of Arabic linguistic thinking see Endress (1986), with a German translation of the essential texts. 2) In the case of Turkic an exception should be made for the Andalusian Arab grammarian Abu Hayyan al-Garnati (d.745/1344). This grammarian wrote extensively about Arabic grammar and Qur'anic exegesis, but he also composed grammars of Ethiopian, Mongolian and Turkic, applying the model of Arabic linguistic theory to these languages (see Ermers 1989). Later grammarians of Turkic were native speakers of the language, but in most instances they followed his example in the analysis of their language. 3) It also occurs in an abbreviated form, al-mudmar, e.g., Kitdb I, 42.13; for an extensive analysis of the terminology involved, see Mosel (1975:102ff.). 4) It is not known who introduced this abstract notation into Arabic grammar. The representation of the radicals of the word with the three consonants /-'-/, to which the auxiliary consonants may be added, is highly suitable for Arabic with its Semitic system of triliteralism. The grammarian az-Zaggagî (d. 340/951) reports an early discussion in which the grammarian Mu'ad al-Harra' (d. 187/802) participated (Magdlis 190.7-8). In this discussion Mu'ad asks someone to derive from a given verb yâ fâ'ilu f'al, i.e., an active participle of the form fâ'il connected with an imperative of the form if'al. According to some sources Mu'ad was the inventor of the science of tasrif, i.e., morphology (cf. Abbott 1972:6). Obviously, the use of an abstract notation to indicate the pattern of a word is a clear sign of awareness of an underlying level in morpho(no)logy. It is, perhaps, worth mentioning that the/-'-/ notation was also used in metrical analysis, by al-Halil (d. 175/791), the author of the first Arabic dictionary, who is generally regarded as the 'inventor' of metrics. As far as I know, however, he does not use the word taqdîr in his Kitdb al- 'ayn. 5) Actually, the example used here is fairly interesting, because it demonstrates that the f-'-l notation may also be used to indicate a change in the order of the consonants. The word in
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
291
question is 'asyd'u, the plural of say' "thing". According to al-Mubarrad the expected plural is fa'lâ'u, i.e., *say'â'u, but in order to avoid a combination of two hamza's with an alif (in the phonological interpretation of â as a + alif), the second hamza — the / consonant of the word — is put in front, and the pattern becomes laf'â 'u. 6) A spectacular example is a paraphrase of the sentence ad-ddriba s-sdtima l-mukrima lmu 'tiyahu dirhaman al-qd 'imu fî dârihi ahuka sawtan akrama l-dkilu ta 'dmahu gulâmuhu zaydun 'amran hâlidin bakran 'abdallâhi ahuka "the person, whose slave ate his food, honoured the person who hit the person who reviled the person who honoured the person to whom the person in whose house your brother is staying gave a dirham, with a whip, namely Zayd, 'Amr, Hâlid, Bakr, 'Abdallâh, your brother". My English translation follows the paraphrase given by al-Mubarrad, which he calls the taqdîr of the original sentence; the proper names at the end indicate with their case-endings their function in the sentence. Note that the clause ad-ddriba ... sawtan is said to have become one noun, sâra isman wâhidan, which serves as the object to akrama. 7) It should be mentioned that there is a possibility that in later Arabic grammar things may have changed. It would be wrong to base one's self on only one quotation and in general I am skeptical about parallels, but I must confess that I find the following quotation from as-Sirbîni rather suggestive. In describing the passive in Arabic this grammarian from the 16th century saysfa-in kâna l-fi'l mâdiyan wa-aradta an tanqula l-maf'ûl bihi min bâb al-maf'ul ilâ hâdâ lbâb 'amiltafihi talâtata a 'mâl "if the verb is perfect and you wish to transfer the object from the category of object to that [of agent] you perform three operations" (Carter 1981:170.1-2). The author then goes on to describe in detail the derivative chain through which the active sentence becomes a passive sentence (cf. also Carter 1981:390.6 on the tamyîz construction garastu larda sagaran, which was also dealt with by Ibn Ginnî; here the author even uses the term muhawwal, which is as near as you can get to the term 'transformed', cf. Carter's comment 20.11, n.3). 8) On the other hand, I believe that Sluiter goes too far when she criticizes Householder for misinterpreting Apollonius' analysis of sentences (1990:67, n.l 13). Her criticism turns on those cases where the underlying structure of a sentence in Apollonius' syntax is ungrammatical (e.g., cases like boulomai emautonperipatefn "I wish me to walk"). In Householder's view, this use of ungrammatical sentences shows that the underlying structure can hardly be an interpretive paraphrase, since it does not contribute to any better understanding of the sentence's meaning. I believe that he is right and refer to Sîbawayhi's use of the phrase lâ yutakallamu bihi "it is not spoken" in those instances where he wishes to give a structural description of a sentence. Basically, Apollonius is doing the same thing: he wishes to make absolutely clear what the structural analysis of the surface sentence is, and the resulting ungrammatical sentence should be regarded as a notation in metalanguage.
REFERENCES A) Primary sources Abu 'Ubayda, Magâz = Abû 'Ubayda Ma'mar ibn al-Mutannâ at-Taymî, Magâz al-Qur'ân. Ed. by Muhammad Fu'âd Sazgin [Fuat Sezgin]. 2 vols. Cairo: Muhammad Sâmî Amîn al-Hângî, 1954.
292
KEES VERSTEEGH
Farrâ', Ma'ânî - Abû Zakariyyâ' Yahyâ ibn Ziyâd al-Farrâ', Ma'ânî lQur'ân. Ed. by Muhammad 'Alî an-Nağğâr. 3 vols. Cairo: ad-Dâr alMisriyya, 1955-72. Ibn Ginnî, Hasâ'is = Abu 1-Fath 'Utman Ibn Ginnî, al-Hasais. Ed. by Muhammad 'Alî an-Naggar. 3 vols. Cairo, 1952-56. (Repr., Beirut: Dar alHuda, n.d.) Ibn al-Muqaffa', Risdla = 'Abdallah Ibn al-Muqaffa', Risalafi s-sahdba. Ed. and transl. by Charles Pellat. Paris: Maisonneuve & Larose, 1976. Ibn as-Sarrag, Usûl = Abu Bakr ibn as-Sari Ibn as Sarrag, Kitdb al-usulfinnahw. Ed. by 'Abd al-Husayn al-Fatli. 3 vols. Beirut: Mu'assasat arRisala, 1985. Mubarrad, Muqtadab = Abu 1-'Abbas Muhammad ibn Yazîd al-Mubarrad, alMuqtadab. Ed. by Muhammad 'Abd al-Haliq 'Udayma. 4 vols. Cairo: Dar at-Tahrîr, 1965-68. Muqatil, Tafsîr = Abu 1-Hasan Muqatil ibn Sulayman al-Balhi, at-Tafsîr. Ed. by 'Abdallah Mahmud Sihata. 4 vols. Cairo: al-Hay'a al-Misriyya al'Amma li-1-Kitab, 1980-87." Sîbawayhi, Kitdb = Abu Bisr 'Amr ibn 'Utman Sîbawayhi, al-Kitdb. 2 vols. Bulaq, 1316 A.H. (Repr., Baghdad: al-Muthanna Library, n.d.) Zağğâğî, Mağâlis = Abû 1-Qasim 'Abd ar-Rahman ibn Ishaq az-Zağğâğî, Magdlis al-'ulamd'. Ed. by 'Abd as-Salam Muhammad Harûn. Kuwait: Wizarat al-Irsad wa-1-Anbâ', 1962. B) Secondary sources Abbott, Nabia. 1972. Studies in Arabic Literary Papyri. Vol. II: Language and literature. Chicago & London: Univ. of Chicago Press. Ayoub, Georgine. 1990. "De ce qui 'ne se dit pas' dans le Livre de Sibawayhi: La notion de tamtil". Studies in the History of Arabic Grammar II ed. by Michael G. Carter & Kees Versteegh, 1-15. Amsterdam & Philadelphia: John Benjamins. Blank, David L. 1982. Ancient Philosophy and Grammar: The Syntax of Apollonius Dyscolus. Chico, Calif.: Scholars Press. Bohas, Georges. 1981. "Quelques aspects de 1'argumentation et de l'explication chez les grammairiens arabes". Arabica 28.204-221. Brunschvig, Robert. 1970/72. "De la fiction legale dans l'Islam medieval". Studia Islamica 32.41-51. Carter, Michael G. 1968. A Study of Sibawaihi 's Principles of Grammatical Analysis. Ph.D. diss., Univ. of Oxford. . 1981. Arab Linguistics: An introductory classical text with translation
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
293
and notes. Amsterdam: John Benjamins. . 1991. "Elision". Proceedings of the Colloquium on Arabic Grammar, Budapest, 1-7 September 1991 ed. by Kinga Devenyi & Tamas Ivanyi, 121-33. Budapest: Eötvos Lorand Univ. & Csoma de Koros Society. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press. . 1966. Cartesian Linguistics: A chapter in the history of rationalist thought. New York & London: Harper & Row. Devenyi, Kinga. 1991. Sîbawayhi and al-Farra: Uniformity and diversity in the history of Arabic grammar-writing. Kandidatusi disszertacio, Univ. of Budapest. Endress, Gerhard. 1986. "Grammatik und Logik: Arabische Philologie und griechische Philosophie im Widerstreit". Sprachphilosophie in Antike und Mittelalter ed. by Burkhard Mojsisch, 163-299. Amsterdam: B. R. Gruner. Ermers, Rob. 1989. "Abu Hayyân al-Andalusi en zijn grammatica van het Turks: Een introductie". Sharqiyyât 2.302-333. Gimaret, Daniel. 1980. Theories de Tacte humain en theologie musulmane. Paris: J. Vrin; Leuven: Peeters. Gruntfest, Yaakov. 1984. "Medieval Arabic Grammarians: First transformationalists?". Zeitschrift der deutschen morgenlandischen Gesellschaft 134. 226-236. Guillaume, Jean-Patrick. 1981. "Le statut des representations sous-jacentes en morphophonologie d'apres Ibn Ginnî". Arabica 28.222-241. Gully, Adrian. 1991. Aspects of Semantics, Grammatical Categories and Other Linguistic Considerations in Ibn-Hishdm's Mughnî al-Labib. Ph.D. diss., Univ. of Exeter. Householder, Fred W. 1981. The Syntax of Apollonius Dyscolus. Translated and with commentary. Amsterdam: John Benjamins. Kofler, Hans. 1933. "Handbuch des islamischen Staats- und Verwaltungsrechtes von Badr ad-Din Ibn Gama'ah, herausgegeben, ubersetzt und mit Anmerkungen versehen". Islamica 6.349-414. Mosel, Ulrike. 1975. Die syntaktische Terminologie bei Sîbawaih. Doctoral Diss., Univ. München. Owens, Jonathan. 1988. The Foundations of Grammar: An introduction to Medieval Arabic grammatical theory. Amsterdam & Philadelphia: John Benjamins. Peled, Yishai. 1990. "Non-Referential Pronouns in Topic Position in Medieval Arabic Grammatical Theory and in Modern Usage". Zeitschrift der deutschen morgenlandischen Gesellschaft 140.3-27.
294
KEES VERSTEEGH
. 1992. "Cataphora and taqdir in Medieval Arabic Grammatical Theory". Jerusalem Studies in Arabic and Islam 15.94-112. Sluiter, Ineke. 1990. Ancient Grammar in Context: Contributions to the study of Ancient linguistic thought. Amsterdam: VU Univ. Press. Talmon, Rafael. 1993. "Two Early 'non-Sîbawaihian' Views of 'amal in Kernel-Sentences". Zeitschrift fur arabische Linguistik 25.278-288. Versteegh, Kees. 1983. "Arabic Grammar and the Corruption of Speech". Arab Language and Culture ed. by Ramzi Baalbaki, 117-138. Beirut: American Univ. of Beirut. . 1990. "Freedom of the Speaker? The term ittisâ' and related notions in Arabic grammar". Studies in the History of Arabic Grammar II ed. by Michael G. Carter & Kees Versteegh, 281-293. Amsterdam & Philadelphia: John Benjamins. . 1993. Arabic Grammar and Qur'ânic Exegesis in Early Islam. Leiden: E. J. Brill. ABSTRACT Even 'naive' speakers use a distinction between actual, realized speech with its 'literal' meaning, and an underlying level of 'what is actually meant'. Such a distinction is made because speakers instinctively feel that very often actual speech does not represent exactly what the speaker intends to say. In this paper it is claimed that this non-technical distinction lies at the basis of a technical distinction between a surface structure of speech and an underlying level. In the technical stage of Arabic grammar the emphasis shifts from an analysis of the underlying intention of the speaker towards an explanation of the syntactic form of actual speech, which is mapped onto an underlying representation. Both in the Classical Greek and the Arabic/Islamic tradition we find a development from an early stage of exegetical activity, in which the intention of the speaker or the text is elaborated by positing an underlying level of semantic representation, towards a technical distinction between a surface level and an underlying level. The difference between the two traditions lies in the fact that Greek linguistics was more semantically oriented, whereas in Arabic grammar the main tool of the grammarians, the taqdîr, was basically an instrument to explain the syntactic structure of speech, in line with the predominantly formal approach of the Arabic grammarians. Compared with modern linguistic theory, both traditions have in common that they do not look for an underlying level of meaning that is universal to all languages. The main reason for this difference is that neither Greek nor Arabic linguists were interested in the study of other languages.
'UNDERLYING LEVELS' IN ARABIC GRAMMATICAL TRADITION
295
RÉSUMÉ Même des locuteurs 'naïfs' ont l'habitude de distinguer entre langage actuel, réalisé avec une signification 'littérale', et un niveau sous-jacent de 'ce que l'on veut dire'. Cette habitude s'explique par le sentiment intuitif que très souvent le langage actuel ne représente pas exactement ce que le locuteur veut dire. Dans notre article nous affirmons que c'est cette distinction nontechnique qui est a la base de la distinction technique entre structure de surface et structure sous-jacente. Chez les techniciens de la grammaire arabe, on n'étudie plus la signification sous-jacente, mais on se dédie à l'explication de la structure syntaxique du langage, qui est projetée sur une représentation sous-jacente. Tant dans la tradition linguistique grecque que dans la tradition linguistique arabo-islamique 1'analyse linguistique commence avec une période d'activité exégétique, pendant laquelle les commentateurs reconstruisent l'intention du locuteur ou du texte par 1'intermédiate d'un niveau sous-jacent de représentation sémantique, et puis, se développe vers une distinction technique entre un niveau de surface et un niveau sous-jacent. Mais il y a là une différence entre ces deux traditions: tandis que les grammairiens grecs s'orientent surtout vers 1'analyse sémantique, les grammairiens arabes utilisent leur instrument principal, le taqdîr, avant tout pour l'explication de la structure syntaxique du langage, conformément à la nature plutôt formelle de leurs recherches. Lorsqu'on compare les deux traditions ensemble avec la théorie linguistique moderne, il devient clair cependant que ce qu'ils ont en commun, c'est l'exclusion d'une orientation universelle dans 1'analyse grammaticale. Ceci s'explique par l'exclusivité de leurs intérêts linguistiques, qui sont monopolisés par 1'étude d'une seule langue.
ZUSAMMENFASSUNG Sogar 'naive' Sprecher kennen einen Unterschied zwischen der wortlichen Bedeutung der aktualisierten Sprache einerseits und der wirklichen Bedeutung auf einer tieferen Ebene andererseits. Sie machen diesen Unterschied, weil sie sich bewußt sind, daß die realisierte Sprache sehr oft nicht der Intention des Sprechers entspricht. In diesem Aufsatz behaupten wir, daB dieser nicht-technische Unterschied dem technischen Unterschied zwischen Oberflächenstruktur und Tiefenstruktur zugrundeliegt. Zum Zeitpunkt, da die arabische Grammatik sich als eine technische Disziplin herausbildete, hatte
296
KEES VERSTEEGH
sich das Interesse der Sprachwissenschaftler von der Analyse der zugrundeliegenden Intention des Sprechers allmahlich der Erklarung der syntaktischen Form zugewandt. Sowohl die griechische wie die arabisch/islamische Tradition machten eine Entwicklung durch von einem frühen Stadium, in dem die Intention des Sprechers oder des Textes auf Grund der semantischen Ebene eruiert wurde, in Richtung eines technischen Unterschiedes zwischen Oberflächenstruktur und Tiefenstruktur. Der Unterschied zwischen den beiden Traditionen liegt jedoch darin, daß die griechische Sprachwissenschaft semantisch orientiert war, wahrend in der arabischen Tradition die Sprachwissenschaftler vor allem die Methode des taqdîr benutzten, die im wesentlichen dazu diente, die syntaktische Struktur der Sprache zu erklaren, ganz im Sinne ihres üblichen Formalismus. In Vergleich mit der modernen Linguistik ist diesen beiden Traditionen gemein, daß sie nicht auf der Suche nach einer universellen semantischen Ebene für alle Sprachen sind: weder die griechischen noch die arabischen Sprachwissenschaftler interessierten sich fur das Studium fremder Sprachen.