“ the verb which ‘ passes over ’ through a preposition ” ) in Medieval Arabic Grammatical Tradition

Contrary to the categorization of verbs with regard to their taʿaddin which modern scholarship has customarily ascribed to the medieval Arab grammarians, the term al-fiʿl al-mutaʿaddī bi-ḥarf jarr is generally not regarded by these grammarians as a subcategory of al-fiʿl al-mutaʿaddī. Furthermore, Arab grammarians do not restrict the application of the term al-fiʿl al-mutaʿaddī bi-ḥarf jarr to constructions in which the prepopositions in question are governed; this has far-reaching repercussions on the notion of ẓarf. The grammarians’ conception of al-fiʿl al-mutaʿaddī bi-ḥarf jarr, surveyed in this article, is explained both against the backdrop of the early transformations the term taʿaddin underwent, and within the grammarians’ general theoretical framework.


Introduction
The following categorization of verbs according to their transitivity is commonly ascribed to the Arab grammarians:2 all verbs belong either to the category of al-fiʿl al-mutaʿaddī (transitive verbs) or to that of al-fiʿl al-lāzim or al-fiʿl ġayr al-mutaʿaddī (also: al-fiʿl alqāṣir) (intransitive verbs); transitive verbs are further subcategorized into al-fiʿl al-mutaʿaddī bi-nafsi-hi (verbs which are transitive through themselves, thereby assigning the accusative to an object) and al-fiʿl al-mutaʿaddī bi-ġayri-hi (verbs which are transitive through something else) or al-fiʿl al-mutaʿaddī bi-ḥarf jarr 3 (also: bi-ḥarf jārr, bi-ḥarf xafḍ, bi-ḥarf xāfiḍ and bi-ḥarf ʾiḍāfa) (verbs which are transitive through a preposition). 4This division will henceforth be referred to as 'the standard categorization'.WRIGHT (1896-1898: II, 46) illustrates al-fiʿl al-mutaʿaddī bi-nafsi-hi with the sentence balaġa-nī l-xabaru ('the news reached me'), and al-fiʿl al-mutaʿaddī bi-ḥarf jarr with the sentence qadara ʿalā šayʾin ('he was able to do something').It is implied that the latter category consists of verbs which govern5 the prepositions in question, to the exclusion of prepositions which are not governed, such as fī in the sentence kataba zayduni l-kitāba fī l-dāri ('Zayd wrote the book in the abode').
These two traits ascribed to the category of al-fiʿl al-mutaʿaddī bi-ḥarf jarr, namely that it is a subdivision of al-fiʿl al-mutaʿaddī, and that its application is restricted to constructions in which the prepositions are governed, will be questioned in Sections 2 and 3, respectively.It will become evident that these two issues are interdependent.
2. The status of al-fiʿl al-mutaʿaddī bi-ḥarf jarr 2.1 Sībawayhi differs significantly from his followers with regard to the way he uses the term taʿaddin (lit."passing over", viz. the subject).As LEVIN (1979: 209) has already noted, Sībawayhi applies the term taʿaddin to the relationship between verbs and accusative nominals only. 6Occasionally, Sībawayhi regards the taʿaddin relationship as excluding prepositional phrases. 7For example, he explains (al-Kitāb, I, 133) the unacceptability of *wahabtu-ka (in the sense of 'I gave you [something]'), by contrast with ʿadadtu-ka ('I  counted [something] for you'), kiltu-ka ('I measured [something] for you') and wazantu-ka ('I weighed [something] for you'), as follows: li-ʾanna-hum lam yuʿaddū-hu ('since they [sc.the Bedouins] do not make it engage in a taʿaddin relationship'); the acceptable construction is, therefore, wahabtu la-ka. 8Moreover, Sībawayhi uses the term taʿaddin only when the accusative nominal in question is a mafʿūl (here in the sense of "direct object", but see KASHER 2012), to the exclusion of, for example, a ẓarf (a locative/temporal qualifier). 9ībawayhi uses the root w-ṣ-l (connect, reach) 10 and the verb ʾaḍāfa (connect, join) 11 to designate the relationship between verbs and prepositional phrases.In an example of the use of the former, he states (al-Kitāb, I, 12) that verbs such as ixtāra ('he chose') are (basically) ʾafʿāl tūṣalu bi-ḥurūf al-ʾiḍāfa ('verbs connected/reaching through words/morphemes not reach it [sc.the mafʿūl] except through the word/morpheme of connection [sc.bi-]').This wording, according to the context, emphasizes the distinction between the two categories: the verb is not connected to (or: does not reach) zayd, except by means of the preposition. 19Unlike the term al-fiʿl al-mutaʿaddī bi-ḥarf jarr, this 'negative' formulation does not necessarily imply that this category is to be subsumed under al-fiʿl al-mutaʿaddī; it may in fact imply the opposite.The importance of the distinction between the 'positive' and the 'negative' wordings will become manifest presently.

2.2
The roots w-ṣ-l and ḍ-y-f are also used in post-Mubarradian writings with reference to the relationship between verbs and prepositional phrases. 20However, already in Ibn al-Sarrāj one encounters a frequent usage of al-fiʿl al-mutaʿaddī bi-ḥarf jarr.Since Ibn al-Sarrāj is known for his systematization of grammar, 21 it makes eminent sense to ask whether the standard categorization is in agreement with his theory, or more specifically, whether for him al-fiʿl al-mutaʿaddī bi-ḥarf jarr constitutes a subcategory of al-fiʿl al-mutaʿaddī.
('If you had said marartu zaydan … and elided the preposition, this would not have been allowed, other than as poetic license.Yet, the preposition [lit."the operator of the genitive", sc.This statement might suggest that Ibn Jinnī mentions this category not for its own sake, but rather as part of his discussion of the accusative nominals, to wit, the (extremely restricted) possibility of using these verbs without a preposition, so that they take accusative nominals, as well as the functioning of such prepositional phrases as if they were accusative nominal. 28Third, Ibn Jinnī does not mention any other category of verbs with regard to prepositional phrases; he does not even mention the category of verbs such as ixtāra (as al-Zajjājī does); the entire discussion following the quoted statement is restricted to accusative nominals.
To recapitulate, although al-fiʿl al-mutaʿaddī bi-ḥarf jarr is regarded as a distinct category for al-Zajjājī and Ibn Jinnī, it is nevertheless marginal vis-à-vis al-fiʿl al-mutaʿaddī bi-nafsi-hi; in contrast to the comprehensive categorization of verbs with regard to the number and type of accusative nominals they take, these grammarians (as well as others) put forward no parallel discussion of prepositional phrases. 29

2.4
As for the other grammarians studied for this paper, we have found in their writings no categorization of verbs according to their taʿaddin which includes al-fiʿl al-mutaʿaddī biḥarf jarr.Al-Jurjānī (al-Muqtaṣid, 595), for instance, defines [al-fiʿl] al-mutaʿaddī as mā naṣaba mafʿūlan bi-hi ('what causes an object to take the accusative'), whereas al-fiʿl ġayr al-mutaʿaddī is defined as mā lam yanṣib mafʿūlan bi-hi ('what does not cause an object to take the accusative').More explicitly, al-ʾAstarābāḏī (Šarḥ, IV, 135-136) holds that verbs such as qaruba ('he was/became near') are regarded as mutaʿaddiya bi-l-ḥarf al-fulānī ('engaging in a taʿaddin relationship through such-and-such a particle'), whereas the term al-mutaʿaddī used without any expression restricting its meaning (ʾiḏā ʾuṭliqa) 30 does not apply to them; they are thus lāzima.That is, taʿaddin by means of a preposition is a trait, or an option, of verbs which are deemed lāzim, in contradistinction to verbs which are categorized as mutaʿaddin. 312.5 That the term taʿaddin, not only for Ibn al-Sarrāj but also for later grammarians, is basically applied with reference to the relationship between verbs and accusative nominals, is manifest also in the grammarians' discussions of the verb daxala.The problem which this verb poses is that the constructions daxala l-bayta and daxala fī l-bayti are both acceptable.32However, the problem as it is formulated by the grammarians is whether the verb is to be categorized, basically, as mutaʿaddin or as ġayr mutaʿaddin (also: lāzim).Ibn al-Sarrāj (al-ʾUṣūl, I, 170), for example, says: wa-qadi xtalafa l-naḥwiyyūna fī daxala lbayta hal huwa mutaʿaddin ʾaw ġayru mutaʿaddin ('grammarians are at variance about whether or not [daxala in] daxala l-bayta is mutaʿaddin').The basic construction is, according to Ibn al-Sarrāj and other grammarians, daxala fī l-bayti (or: ʾilā l-bayti), from which the construction daxala l-bayta is derived by eliding the preposition.The evidence which grammarians adduce for this end aims at proving that this verb is to be categorized as fiʿl ġayr mutaʿaddin.For example, al-Jurjānī (al-Muqtaṣid, 600) notes that the infinitive of this verb is duxūl, of the pattern fuʿūl, which, as he states, is characteristic of verbs which are ġayr mutaʿaddin. 33n a similar vein, Ibn al-Sarrāj (al-ʾUṣūl, III, 86 ff.) categorizes verbs of Form I as either fiʿl mutaʿaddin ʾilā mafʿūl or fiʿl ġayr mutaʿaddin with regard to their patterns (including their cognates); the verb qaʿada, for instance, is subsumed under the latter, with no mention of the fact that it can take a prepositional phrase (see §2.2).34 2.6 Sentences such as marartu bi-zaydin should thus be analyzed as constructed in two stages.At the first stage, marartu is constructed, the verb being ġayr mutaʿaddin, since it takes no accusative nominal.At the second stage, the verb nevertheless engages in a taʿaddin relationship with some constituent, but only by means of a preposition, since it cannot take accusative objects.This two-stage model is inferred from the above discussion, and is explicit in several passages in al-Jurjānī; for example, this grammarian says (al-Muqtaṣid, 699): … ʾanna-ka taqūlu marartu fa-lā yataʿaddā fa-taʾtī bi-l-bāʾi35 fa-taqūlu marartu bi-zaydin fa-tūṣilu l-bāʾu l-fiʿla ʾilā zaydin … ('that you say marartu and it does not engage in a taʿaddin relationship [lit.does not "pass over"].And then you use the bi-and say marartu bi-zaydin, and the bi-connects the verb with/makes the verb reach zayd …'). 36tional phrase is an option open for verbs generally.
Grammarians occasionally explain that such verbs are too 'weak' to take an accusative nominal, and therefore require the aid of a preposition. 37For instance, Ibn Jinnī, discussing such constructions (al-Xaṣāʾiṣ, I, 341), says: marartu bi-zaydin wa-mā kāna naḥwa-hu mim-mā yalḥaqu min ḥurūfi l-jarri maʿūnatan li-taʿaddī l-fiʿli ('marartu bi-zaydin and similar [constructions containing] prepositions joining in order to aid the taʿaddin of the verb'). 38To this theory belongs also the use of the term taʿdiya, which means: causing a verb to be mutaʿaddin. 39Grammarians occasionally state that there are three elements which have the effect of taʿdiya: Form IV, Form II and the preposition bi-. 40This process has not only the effect of increasing the number of constituents, but also of 'causativization'. 41However, it is sometimes stated that the process of taʿdiya is not restricted to the preposition bi-, since taʿdiya is a function of all prepositions.42For instance, al-Jurjānī (al-Muqtaṣid, 346ff.)discusses the three methods of causing a verb which is ġayr al-mutaʿaddī to be al-mutaʿaddī: al-hamza, i.e.Form IV, al-bāʾ and al-taḍʿīf, i.e.Form II, the second being illustrated with the construction ḏahaba ʿamrun bi-zaydin ('ʿAmr went away with Zayd' or 'ʿAmr made Zayd go away'); however, al-Jurjānī also remarks: wakaḏā jamīʿu ḥurūfi l-jarri ('and this applies to all prepositions'), a statement which he illustrates with the construction ḏahabtu ʾilā zaydin ('I went to Zayd') and xarajtu mina lbaṣrati ('I went out of Baṣra'). 43This corroborates our conclusion that taking prepositional phrases is not regarded by the grammarians as an inherent trait by which verbs are categorized, in contrast to accusative nominals.

2.7
This function implemented by prepositions, namely creating a syntactic link between verbs and nominals to which they do not assign the accusative, has a semantic aspect as well: in the constructions in which they appear, prepositions impose certain meanings,44 to which grammarians dedicate lengthy discussions. 45For example, al-Zajjājī (Ḥurūf, 47) says that in the construction marartu bi-zaydin, the preposition bi-imposes the meaning of ʾilṣāq ('attachment'), so that at the second stage (after this verb is regarded, at the first stage, as ġayr mutaʿaddin), the preposition bi-not only enables the verb to take the constituent at stake, but also determines the semantic relationship between the verb and this constituent, by imposing a certain meaning.
2.8 This is not to say, however, that every token of a preposition implements the function of connecting a verb (or a verb-like element) to a nominal.Grammarians hold that prepositions which are deemed otiose (zāʾid, mazīd or ziyāda) do not implement this function. 46Moreover, Ibn Hišām (Muġnī, 493) states that exceptive prepositions, e.g.xalā, do not cause taʿdiya, due to their semantic significance: they do not connect the meanings of verbs with the following nominals, since they preclude the semantic connection between verbs to nominals which is a characteristic of prepositions causing taʿdiya.47 3. Restrictions on taʿaddin with respect to government 3.1 The grammarians frequently parse prepositional phrases, or their genitive nominal only, 48 as mafʿūl (bi-hi), and regard them as (virtually) taking the accusative49 case. 50For example,I,145) accounts for the syncretism between the accusative and the genitive forms of the sound pluralboth the feminine and the masculine (e.g.muslimātin ('Muslim women') and muslimīna ('Muslim men'), respectively)and between those of the dual (e.g.rajulayni 'two men'), first by pointing to the fact that this applies also to personal pronouns, 52 and second by claiming that both the accusative and the genitive nominals are mafʿūl, since the meaning of marartu bi-zaydin (representing here the genitive) is: faʿaltu 53 hāḏā bi-hi ('I did this [act] to him').That is, since Zayd is the one to whom the action (designated by the verb) was done, zayd is the mafʿūl bi-hi.54 In other words, the term mafʿūl (bi-hi) applies not only to direct objects, but also to genitive nominals introduced by prepositions, since they engage in the same semantic relationship with the verb.55 al-ʾAstarābāḏī (Šarḥ, I, 333-334, IV, 135-136)   idem, al-Lumaʿ, 14, 22; al-Jurjānī, al-Muqtaṣid, 234-235, 335, 353, 376-377, 613, 781, 814, 824, 830,  851, 1080; Ibn al-ʾAnbārī, ʾAsrār, 52, 69; al-ʾAstarābāḏī, Šarḥ, I, 62-63, 183, 204, 219, 221, 244, 503,  II, 102, 190, 302, 425, III, 465, IV, 137, 261-262, 289, 327; Ibn Hišām, Muġnī, 465, 520-521; idem,  Šarḥ, 267.This is inferred also from Sībawayhi, al-Kitāb, I, 393; Ibn Jinnī, al-Xaṣāʾiṣ, I, 397; al-Zamaxšarī, al-Mufaṣṣal, 343.See OWENS 1988: 176-177.On the status of prepositional phrases as oneword equivalents, see BAALBAKI 1999: 93 ff.Note, however, that prepositional phrases are a marked mafʿūl: Ibn Jinnī (al-Xaṣāʾiṣ, I, 397), for instance, while analyzing a poetic verse which includes both a nominal mafʿūl and a prepositional phrase, refers to the former, in contradistinction to the latter, as a mafʿūl bi-hi ṣaḥīḥ (i.e."proper"). AlʾAstarābāḏī (Šarḥ, I, 333-334) holds, moreover, that the term mafʿūl bi-hi, as long as it is muṭlaq and not muqayyad, i.e. restricted by a restrictive expression (for the pair muṭlaq vs. muqayyad, see Levin 1991: 920 ff., and see §2.4), does not refer to prepositional phrases.Elsewhere this grammarian refers to a prepositional phrase as a naṣb which is not ṣarīḥ (Šarḥ,  I, 195). Se also ibid, III, 409, IV, 135-136, 265; Ibn Hišām, Muġnī, 465.An even more radical distinction is found where Ibn Hišām (ibid, 152) adduces a poetic verse from which a preposition is elided, and in consequence the genitive nominal becomes, according to him, mafʿūl.Note that it is this inclusion of prepositional phrases under mafʿūl bi-hi which leads al-ʾAstarābāḏī to his dissenting theory, which will be discussed in §3.5.51 This correspondence is, however, not without exception, e.g.where grammarians explicitly distinguish between direct objects and prepositional phrases on semantic grounds, see Ibn al-Sarrāj, al-ʾUṣūl, I,  171; Al-Jurjānī, al-Muqtaṣid, 599-602, 613-614.This can also be inferred from al-Mubarrad, al-Muqtaḍab, II, 271.52 The illustration adduced here is that the form of the second person masculine singular is -ka, both in the accusative and genitive, e.g.raʾaytu-ka ('I saw you') and marartu bi-ka ('I passed by you'), respectively.53 It is also plausible to read here: faʿalta, as if the grammarian is addressing the person performing the utterance marartu bi-zaydin. 54The phrase mafʿūl bi-hi here constitutes what PELED (1999) labels a 'metagrammatical intuitive term', i.e. "… its semantic composition comprises components from the grammatical as well as from the nongrammatical everyday concept underlying it" (ibid: 58).55 The same argument recurs in al-Zajjājī, al-ʾĪḍāḥ, 128.See also al-ʾAstarābāḏī, Šarḥ, II, 425.Interestingly enough, al-Jurjānī (al-Muqtaṣid, 186) accounts for these morphological phenomena by putting the accusative and the genitive nominals on a par (in a discussion resembling al-Mubarrad's in its wording), but nevertheless the equivalence for him lies in the fact that both the direct object, e.g. in ḍarabtu zaydan ('I hit Zayd'), and the prepositional phrase, e.g. in marartu ilā l-baṣrati ('I went to Baṣra') (later in this discussion he also mentions marartu bi-), are faḍla, that is, they are dispensable constituents: both ḍarabtu and marartu qualify as sentences, in contrast to verbs lacking a nominative nominal (i.e. a subject).Ibn al-ʾAnbārī (ʾAsrār, 22-23) accounts for the syncretism in the dual and the sound masculine plural by alluding to five or six factors; these include the dispensability of both the accusative and geni-states that a purely semantic definition of the terms mafʿūl bi-hi and mutaʿaddin 56 entails parsing prepositional phrases as mafʿūl bi-hi and regarding verbs which do not take a direct object as mutaʿaddin.Although he himself does not adhere prima facie to a purely semantic criterion, he states that … wa-l-taʿaddī wa-l-luzūmu bi-ḥasabi l-maʿnā ('taʿaddin vs. lack of it are determined according to the meaning [of the verb in question]').57 The parsing of prepositional phrases as mafʿūls might constitute the flip side of the fact that grammarians regard the relationship existing in such constructions as taʿaddin, although it is found already in Sībawayhi, who does not regard this relationship as taʿaddin (see §2.1).

3.2
The question which now arises is whether this equivalence between prepositional phrases and direct objects is restricted to a defined subcategory of the former, i.e. those which are introduced by governed prepositions.But before tackling this question, some words regarding the concept of 'government' are in order.
There is no universally agreed upon definition of 'government' in this context, i.e. of prepositions governed by verbs, let alone an agreement on exactly which cases should be subsumed under this category.As BADAWI, CARTER and GULLY (2004: 380) state, "[i]ndirectly transitive verbs using prepositions are largely a lexical matter …".That is, a verb's entry in the lexicon includes not only information about the direct object(s) it takes (if any), but also about the preposition(s) which it governs (if it does at all), but not about prepositions introducing adjuncts. 58The relationship between verbs and the prepositions they govern is close, "such that the verb determines the choice of preposition" (HUDDLESTON 1984: 201), in contrast to adjuncts. 59There is one facet, occasionally introduced by modern scholars as a criterion for deciding whether or not a certain preposition is governed by its verb, which is irrelevant as long as one discusses the Arabic grammatical tradition; GLINERT (1989: 151) says that "[q]uite generally, object prepositions are intrinsically meaningless whereas adverbial prepositions are intrinsically meaningful … [b]ut these are just the two extremes of a whole spectrum of meaningfulness in prepositions … ". 60 As we have seen above (in §2.7), however, the Arab grammarians in general assign each preposition a meaning, or several meanings, which it imposes in the constructions in which it occurs, and this applies also to governed prepositions; moreover, many grammarians regard this imposition of meaning as the basic trait of the part of speech 'particle' (viz.ḥarf jāʾa li-maʿnan fī ġayri-hi). 61Regarding obligatoriness (which is another trait occasionally introduced with this respect), since prepositional objects are deemed faḍla, they possess a tive mominals (i.e.their being faḍla), the fact that accusative and genitive personal pronouns share the same form and the semantic equivalence between them.56 As does Ibn al-Ḥājib, the author of the treatise of which al-ʾAstarābāḏī's book constitutes a commentary.57 For further cases where such a semantic correspondence is mentioned, see  BEESTON 1970: 87; HUDDLESTON 1984: 178, 180, 203; GLINERT 1989: 520 (ch.15, n.12).59 See also HUDDLESTON 1984: 177, 224; GLINERT 1989: 153 ff.60 See also HUDDLESTON 1984: 203; GLINERT 1989: 151-152.61 See fn.44.certain degree of dispensability.However, LEVIN (1995) has already shown that the mafʿūl bi-hi occupies an intermediate position between totally indispensable constituents, viz.ʿumda, and totally redundant constituents. 62Several points raised in the following discussion shed some light on the status of prepositional phrases in this respect. 63n the present section, the term 'prepositional object' will designate a prepositional phrase introduced by a governed preposition.I refrain from the term 'indirect object', occasionally used in this meaning, 64 since it is often used to designate nominals associated with the semantic role of recipient. 65For the same reason, I also refrain from using the term 'oblique object'. 66The term 'direct object' is thus used here to designate any accusative nominal object, irrespective of the semantic role it possesses.
It seems that the standard categorization takes into consideration only direct and prepositional objects, to the exclusion of prepositional phrases which are adjuncts.That is to say, the term al-fiʿl al-mutaʿaddī bi-ḥarf jarr cannot apparently be applied, according to the standard categorization, when the preposition in question is not governed.Now, although, as noted, the line separating governed from non-governed prepositions is extremely blurred and depends on one's linguistic theory, the restriction of al-fiʿl al-mutaʿaddī bi-ḥarf jarr to governed prepositions can nevertheless be rejected, even by taking into consideration only clear-cut cases, while excluding problematic cases from the discussion. 67he very fact that the grammarians do not restrict their assertion that prepositions cause taʿdiya (so that the verb in question is mutaʿaddin by means of the preposition in question) by criteria of government, is a very strong argumentum ex silentio.However, there is also explicit evidence for the lack of such criteria, namely instances where grammarians apply the term taʿaddin to prepositions which can by no means be regarded as governed by the verbs.The most clear-cut cases are the following: Al-Jurjānī (al-Muqtaṣid, 715-716) asserts that sentences such as ʾatā-nī l-qawmu ḥāšā zaydin ('the people came to me, except Zayd'), in which ḥāšā is a preposition (ḥarf jarr), are cases of taʿaddin by means of this preposition; being a preposition of exception (istiṯnāʾ), it is not governed by any verb. 68l-Zamaxšarī (al-Mufaṣṣal, 380) refers to the verb which takes ḥattā as al-fiʿl al-muʿaddā bi-hā ('the verb which is made to engage in a taʿaddin relationship through it'). 69l-ʾAstarābāḏī (Šarḥ, I, 189) refers to the prepositional phrase in the sentence qataltu bi-ʾaxī-hi zaydan ('I killed Zayd due to his brother' or 'by means of his brother') as [… mim-]mā yataʿaddā ʾilay-hi l-fiʿlu bi-ḥarfi l-jarri ('with what the verb engages in a taʿaddin relationship [lit."what the verb passes over to"] through a preposition').This bi-conveys the meaning of reason (or, alternatively, an instrumental meaning).
al-ʾAstarābāḏī (ibid, II, 148), in similar fashion, regards the verb qatala as mā ʿuddiya bi-ḥarfi l-jarri ('what is made to engage in a taʿaddin relationship through a preposition') in: al-marʾu maqtūlun bi-mā qatala bi-hi ʾin sayfin fa-sayfin ('a man is killed with what he killed with, if he [killed with] a sword, [he is killed with] a sword').Here it conveys an instrumental meaning.
Finally, al-ʾAstarābāḏī (ibid, IV, 261) regards the preposition li-in the constructions zaydun ʿinda-ka li-ʾikrāmi-ka ('Zayd is at your place out of respect for you') and zaydun fī l-dāri li-ʾikrāmi-ka ('Zayd is in the abode out of respect for you') as muʿaddin li-l-fiʿli lmuqaddari ʾaw li-šibhi-hi ('making the underlying verb or semi-verb be mutaʿaddin'), viz.istaqarra ('he stayed') or mustaqirrun (the active participle thereof), respectively; 70 this preposition, conveying the meaning of purpose, can here by no means be deemed to be governed by the verb istaqarra. 71

3.3
On the other hand, there is no doubt that the grammarians were aware of the fact that certain verbs govern certain prepositions, and that not every preposition can be construed with every verb.We have seen above (in §2.3) that grammarians occasionally categorize verbs as mutaʿaddin bi-ḥarf jarr, where this appellation seems to refer to an inherent trait of the verb to which it is applied.This is manifest, first and foremost, in al-Zajjājī's categorization of verbs according to their taʿaddin; recall that al-Zajjājī differentiates between fiʿl lā yataʿaddā ʾilā mafʿūl and fiʿl lā yataʿaddā ʾillā bi-ḥarf xafḍ, thus regarding fiʿl mutaʿaddin bi-ḥarf jarr as a category by itself.Now, since every verb may take certain prepositions which it does not govern, if the latter category had been inclusive (as it is in Ibn al-Sarrāj, see §2.2), there would have been no reason to differentiate between the two categories.Hence it must consist of only a subcategory of verbs not taking accusative nominals, and it is safe to assume that a certain criterion of government is at stake here.
The notion that certain prepositions occur with certain verbs is most explicitly expressed in Ibn Jinnī's (Sirr, 124) following statement: wa-xuṣṣa kullu qabīlin min hāḏihi l-ʾafʿāli bi-qabīlin min hāḏihi l-ḥurūfi ('every class of these verbs was exclusively assigned with a class of these particles').Note also the following discussions where grammarians allude to this notion: 72 Ibn al-Sarrāj (al-ʾUṣūl, II, 352) states that syndetic relative clauses of the configuration: 70 On the istaqarra/mustaqirrun hypothesis, see PELED 2009: 152-155.71 See also al-ʾAstarābāḏī's (Šarḥ, IV, 289) discussion of rubba, where he regards its relationship with its verb as taʿaddin, if it is classified as a preposition, whereas he himself classifies it as an ism (see also Ibn Hišām, Muġnī, 493).A less clear-cut case is the view mentioned by Ibn Hišām (Muġnī, 111-112), to the effect that the verb masaḥa ('he wiped') engages in a taʿaddin relationship, by means of the preposition bi-, with a constituent referring to the muzīl ('remover'), e.g.water; it is inferred that in such constructions this preposition conveys the meaning of istiʿāna, i.e. an instrumental meaning.72 See also Sībawayhi, al-Kitāb, 12-13.relative pronoun + verb I (+ accusative nominal or prepositional phrase) + conjunction particle + verb II, where the latter verb's accusative nominal or prepositional phrase is covert, are acceptable only if the two verbs are identical with regard to their taʿaddin.It is therefore permissible to say allaḏī ḍarabtu fa-ʾawjaʿtu zaydun ('that whom I hit and pained is Zayd'), since they both take a direct object, the underlying structure of the relative clause being thus allaḏī ḍarabtu-hu fa-ʾawjaʿtu-hu.In the same fashion, it is also permissible to say allaḏī ʾaḥsantu ʾilay-hi wa-ʾasaʾtu zaydun ('that toward whom I acted well and meanly is Zayd'), since both verbs are identical with respect to the preposition they take, viz.ilā.On the other hand, it is impermissible to say allaḏī ḏahabtu ʾilay-hi 73 wa-kafaltu zaydun ('that to whom I went and for whom I was answerable is Zayd'), since the latter takes a different preposition, viz.bi-.This condition is formulated in the following way: ʾiḏā kāna l-fiʿlāni muttafiqayni fī l-taʿaddī wa-fī l-ḥarfi llaḏī yataʿaddayāni bi-hi … ('when the two verbs agree with respect to taʿaddin and the particle through which they engage in a taʿaddin relationshop …').Some notion of government seems to pertain to this discussion.
Sporadic cases in which grammarians refer to prepositions which verbs typically take are ubiquitous.
Interesting are also cases in which grammarians exclude certain verbs from engaging in a taʿaddin relationship with phrases introduced by certain prepositions.For instance, Ibn Hišām (Muġnī, 362) excludes the verb katama ('he concealed') from engaging in a taʿaddin relationship by means of min.The very exclusion of certain verb + preposition combinations from membership in the club of taʿaddin is very strong evidence for its exclusiveness.
Consequently, grammarians find it necessary to explain cases in which verbs take prepositions which they do not regularly take (as an inherent trait they possess, i.e. governed prepositions).Such cases are occasionally accounted for as taḍmīn, 'implication of meaning'. 76Ibn Jinnī (al-Xaṣāʾiṣ, II, 308ff.)maintains, for instance, that when one verb conveys 73 The prepositional phrase is lacking in one of the manuscripts.the meaning of another, the former is sometimes used with the preposition governed by the latter instead of its own (Ibn Jinnī here uses the term taʿaddin with regard to the relationship between the verbs and their respective prepositions). 77Ibn Jinnī refers to the preposition in question as al-ḥarf al-muʿtād maʿa mā huwa fī maʿnā-hu, to wit, the preposition which is habitually used with the verb whose meaning is implied by the verb which takes this preposition in the problematic construction.This is exemplified by the following Qurʾānic verse: uḥilla la-kum laylata l-ṣiyāmi l-rafaṯu ʾilā nisāʾi-kum … (Q 2:187) ('Permitted to you, on the night of the fasts, is the approach to your wives' (ALI 2000: 23)).The verb rafaṯa, says Ibn Jinnī, does not (usually) take the preposition ilā, but either bi-or maʿa. 78The use of ʾilā in this case is accounted for by the semantic correspondence between rafaṯa, in this verse, and ʾafḍā + ʾilā; the close relationship betweenʾ ilā and ʾafḍā is referred to by the statement: bābu-hu l-ʾifḍāʾu ('the "category" to which it belongs is ʾifḍāʾ').The same Qurʾānic verse is discussed in a similar fashion also by Ibn Hišām (Muġnī, 762-764), who adduces it as a case of taḍmīn: ḍummina l-rafaṯu maʿnā l-ʾifḍāʾi.That is, the occurrence of the preposition ilā in this verb is explained by the fact that the verb ʾafḍā governs it, 79 whereas al-rafaṯ basically governs the preposition bi-(wa-ʾinnamā ʾaṣlu l-rafaṯi ʾan yataʿaddā bi-l-bāʾi). 80The following case also points in the same direction.Al-Jurjānī (al-Muqtaṣid, 614-616) opposes the common view that the verb istaġfara ('he asked [someone] to forgive') basically governs the preposition min, which may be elided.Since, for him, istaġfartu [llāha] means saʾaltu llāha ʾan yaġfira ('I asked God to forgive'), in accordance with one of the meanings of Form X (viz.al-ṭalab wa-l-suʾāl 'demanding and asking'), the change in transitivity here is identical with the change in transitivity when using Form IV, whereby only one direct object is added, while all other objects are kept intact.Thus, if istaġfara had taken min, ġafara should also have taken this preposition, which is not the case.In order to account for the (derivative) use of min with istaġfara, al-Jurjānī resorts to the explanatory tool al-ḥaml 81 ʿalā al-maʿnā wa-l-naẓīr ('making [something] accord with [its] meaning and [its] like'): 82 lammā kāna fī-hi maʿnā tubtu wa-ʾanabtu ʿuddiya bi-min ('since it conveys the meaning of tubtu and ʾanabtu [sc.I repented], it is made to engage in a taʿaddin relationship with min').That is to say, since the verb tāba, for example, governs the preposition min, istaġfara, conveying the same meaning, also takes this preposition, by analogy.
Note also that the constructions used by the grammarians in order to illustrate al-fiʿl al-mutaʿaddī bi-ḥarf jarr include prepositions which are governed by the verbs in question (or, at least, may be regarded as such); for instance, one of the most frequent constructions used by grammarians in order to demonstrate their arguments is marra + bi-.
Grammarians also mention cases of ambiguity in verbs, taking into consideration the types of constituent the verbs take in each meaning they convey.For example, al-Mubarrad (al-Muqtaḍab, IV, 96) lists three meanings of the verb wajada: (1) wajadtu ʿalay-hi, 83 derived from the infinitive mawjida, i.e. to be angry with; (2) wajadtu, in the sense of wajadtu l-ḍāllata (taking one direct object), i.e. to find; (3) wajadtu, in the sense of ʿalimtu, i.e. to find out, as in wajadtu zaydan karīman ('I found Zayd noble') (taking two direct objects, as a cognitive verb). 84wo passages in al-ʾAstarābāḏī's treatise have expressions indicating notions which seem to be close to that of government.First, according to this grammarian (Šarḥ, I, 218-219), the subject of a passive verb 85 must be min ḍarūriyyāt al-fiʿl (lit."one of the indispensable things of the verb"), just like the fāʿil (the subject of the active verb) and the mafʿūl of verbs which are categorized as mutaʿaddin.Similarly, he says, the prepositional phrase is mafʿūl bi-hi, though by means of a preposition.From this semantic criterion he concludes that prepositional phrases which are not min ḍarūriyyāt al-fiʿl are not allowed to implement this function, for example prepositional phrases introduced by li-denoting taʿlīl, i.e. indicating the reason, since there are actions which are performed for no purpose.Therefore, from the active jiʾtu-ka li-l-samni ('I came to you for the clarified butter') one cannot derive *jīʾa li-l-samni.Note, however, that the class of constituents fulfilling the condition of min ḍarūriyyāt al-fiʿl is significantly broader than the one dictated by the notion of government, since the former includes also nominals denoting time and place, as well as the maṣdar. 86lsewhere he asserts (ibid, III, 25-26) that it is permissible to elide the resumptive pronoun in syndetic relative clauses where the pronoun follows a preposition (in which case the preposition is also elided), but only if this preposition is muʿayyan (lit."individuated, particularized"), lest the addressee posit a different preposition.The illustrations al-ʾAstarābāḏī adduces for this concept of taʿyīn include ʾamara + bi-('he commanded [someone] + to') and ḥajja + ʾilā ('he performed the ḥajj + to'); for example, the Qurʾānic expression ʾa-nasjudu li-mā taʾmurunā (Q 25:60) ('Shall we adore that which thou commandest us?' (Ali 2000: 299)) means … taʾmurunā bi-hi, hi constituting the resumptive pronoun.The concept of taʿyīn is nevertheless not identical with that of government.First, al-ʾAstarābāḏī states that it applies also to cases such as marartu bi-llaḏī mararta ('I passed by that whom you passed by'), meaning … mararta bi-hi, where a preposition which is not muʿayyan (see in what follows), yet identical with the elided preposition, precedes the relative pronoun in question. 87This concept thus seems to pertain also to the environment in which the verb in question occurs, which is not the case with the notion of government.Second, al-ʾAstarābāḏī adduces the construction allaḏī marartu zaydun, from which the prepositional phrase bi-hi is elided, in order to illustrate his contention that although some prepositions do not fulfill the condition of taʿyīn (wa-ʾin lam yataʿayyan), they nevertheless can be elided along with the resumptive pronouns; the lack of taʿyīn stems from the fact that it is permissible to posit other prepositional phrases here, e.g.maʿa-hu88 and la-hu.The principle of taʿyīn thus entails that the addressee can posit only one possible preposition, whereas a verb may govern more than one preposition.89

3.4
In light of the previous subsection, why did the grammarians not restrict the extension of the term taʿaddin to cover only some of the combinations of verb + preposition, according to a criterion of government?
The main reason seems to be that the term taʿaddin came to acquire an inclusive meaning, and denote (after Sībawayhi) the relationship between verbs and every constituent to which they assign the accusative (overtly or covertly), without any exception (see §2.1).Furthermore, the two-stage model of taʿaddin by dint of prepositions (see §2.6) can also account for the unrestricted application of taʿaddin to prepositions: according to this model, each verb not taking a direct object is basically ġayr mutaʿaddin, and becomes mutaʿaddin only secondarily, by means of a preposition; what is of relevance for the present discussion is that this model applies to all prepositions, irrespective of whether or not they are governed.It thus applies to bi-in marartu bi-zaydin in the same sense that it applies to ḥattā.It also seems that the grammarians' semantic conception of the prepositions (see § §2.7, 3.2) plays a crucial role in this lack of distinction, since prepositions governed by verbs do not differ semantically from those which are not.
There may well be a syntactic reason for the lack of differentiation between prepositions governed by verbs and those which are not, namely, the fact that the syntactic behavior of these two classes appears to be identical, or at least very similar.In other words, if there had been some syntactic feature common to both direct and prepositional objects, but not to prepositional phrases which are adjuncts, there would have been a syntactic basis for singling out prepositional objects from all other prepositional phrases.But since this is not the case, there is no solid syntactic motivation to restrict the extension of taʿaddin in such a way.
Accordingly, grammarians occasionally regard prepositional phrases which are adjuncts as taking (virtually) the accusative case (see §3.1).Thus, al-Jurjānī (al-Muqtaṣid, 1080) regards prepositional phrases introduced by ḥattā as fī mawḍiʿ naṣb, just like those occurring in the sentences ḏahabtu ʾilā zaydin and marartu bi-zaydin.In addition, al-ʾAstarābāḏī (Šarḥ, I, 449) refers to the prepositional phrase in ʾa-bi-l-sawṭi zaydan ḍarabta-hu ('was it with the whip that Zayd [acc.],you hit him?') as al-manṣūb maḥallan. 90hat I would like to show now is that the four syntactic features which prepositional objects are found to share with direct objects are also shared by prepositional phrases which are adjuncts. 911) The syntactic equivalence between prepositional phrases and direct objects is manifest, first and foremost, in constructions displaying a conjunction between a prepositional phrase and a direct object, i.e. constructions of the configuration: verb + prepositional phrase + conjunctive particle + accusative nominal.
These are adduced by grammarians in order to demonstrate the equivalence between prepositional phrases and direct objects.The accusative case of the latter is occasionally explained as stemming from the virtual accusative case of the prepositional phrase; alternatively, grammarians posit an underlying structure in which a verb, conveying the same meaning of the verb in the surface structure, but which is directly transitive, precedes the accusative nominal and assigns it this case.These two explanations are sometimes seen as two sides of the same coin.For example, al-Mubarrad (al-Muqtaḍab, IV, 154) explains the accusative of ʿamr in marartu bi-zaydin wa-ʿamran in the following fashion: li-ʾanna maʿnā-hu ʾataytu fa-ḥamala-hu ʿalā l-maʿnā ʾiḏ kāna qawlu-ka bi-zaydin baʿda marartu fī mawḍiʿi naṣbin ('since its meaning is ʾataytu [sc.I came to], so he [sc.the speaker] made it accord with its meaning, as bi-zaydin following marartu occupies the position of a [nominal taking the] accusative').This embraces the two explanations above: first, the prepositional phrase functions as if it were an accusative nominal, therefore a nominal conjoined to it can also take the accusative; second, an underlying verb, viz.ʾataytu, is posited, whose meaning is similar to the meaning of the verb in the surface structure. 92owever, In order to illustrate this phenomenon, Ibn al-ʾAnbārī (al-ʾInṣāf, 333-335) adduces, inter alia, the following poetic verses: kašḥan ṭawā min baladin muxtārā min yaʾsati l-yāʾisi ʾaw ḥiḏārā ('he departed from a land, preferring this, out of desperation of the desperate or out of caution'), fa-ʾin lam tajid min dūni ʿadnāna wālidan wa-dūna maʿaddin … ('if you find no father after ʿAdnān and Maʿadd) and … ʾiḏā mā talāqaynā mina l-yawmi ʾaw ġadā which has syntactic implications, i.e. with regard to the possibility to function as a subject of a passive verb (see below).Yet, as we have seen, this distinction is far from identical with the one based on government.Moreover, even this grammarian, as we shall see below, on another occasion permits a prepositional phrase which is an adjunct to function as the subject of a passive verb.92 See also Sībawayhi, al-Kitāb, I, 38, 130; al-Mubarrad, al-Muqtaḍab, IV, 111, 153; Ibn al-Sarrāj, al-ʾUṣūl, II, 13-14, 64-65; Ibn Jinnī, al-Xaṣāʾiṣ, I, 102-103, 106-107, 341-343; idem, Sirr, 130-131; Ibn al-ʾAnbārī, al-ʾInṣāf, 327, 331; al-ʾAstarābāḏī, Šarḥ, I, 63, IV, 137, 261.See OWENS 1988: 176-177.For a similar analysis of a similar construction, see Sībawayhi, al-Kitāb, I, 217; Ibn Jinnī, Sirr, 131.Ibn  Hišām (Muġnī, 525-526), however, poses for these constructions the condition that the presence of the preposition in question should not be obligatory.Hence, whereas constructions such as laysa zaydun bi-qāʾimin wa-lā qāʿidan ('Zayd is neither standing nor sitting') are acceptable, since it is permissible to omit the preposition bi-, marartu bi-zaydin wa-ʿamran is unacceptable, for him, since it is impermissible to say *marartu zaydan.
('… when we meet today or tomorrow'), 93 in each of which min is not governed. 94) The equivalence between prepositional phrases and direct objects is also discussed with regard to the ištiġāl (lit."being occupied") phenomenon.95 In the basic ištiġāl constructions an accusative nominal is followed by a verb (in which the subject is incorporated) with an accusative personal pronoun, this latter co-referring with the accusative nominal, e.g.ʿabda llāhi ḍarabtu-hu (ʿAbdallāh [acc.],I hit him').96 This notwithstanding, grammarians discuss also constructions of the configuration: accusative nominal + verb + preposition + genitive personal pronoun (the latter co-referring with the first nominal), e.g.zaydan marartu bi-hi ('Zayd [acc.],I passed by him'). 97Now, with regard to the basic construction the grammarians can easily posit an underlying verb, identical with the verb in the surface structure, preceding the accusative nominal and assigning it its case, i.e. (for the example above) *ḍarabtu ʿabda llāhi ḍarabtu-hu ('I hit ʿAbdallāh, I hit him'). 98Positing an underlying verb identical with the verb occurring in the surface structure is obviously precluded with regard to the second construction, since the verb in the surface structure takes a prepositional phrase, whereas the underlying verb should assign the accusative to the first nominal, this being its raison d'être. 99Grammarians thus posit a different underlying verb, which is semantically identical with (or, at least, akin to) the verb in question, but is mutaʿaddin by means of itself, i.e.
, conveying here the benefactive meaning.Note that al-Jurjānī's (ibid) explanation of this reconstruction is the same as his explanation of zaydan marartu bi-hi, assigning to the verb ʾaʿadda the attribute of mutaʿaddin by means of a preposition: li-ʾanna ʾaʿadda qad taʿaddā ʾilā ḍamīri l-ẓālimīna bi-l-jārri wa-l-fiʿlu l-muḍmaru nāṣibun ('since ʾaʿadda engages in a taʿaddin relationship with the pronoun of [sc.referring back to] al-ẓālimīna through the preposition, while the covert verb assigns the accusative').
Examples of this sort are found already in Sībawayhi (al-Kitāb, I, 42-43): amidst his discussion of such a construction introduced by the interrogative particle ʾa, one can find the following two sentences: ʾāl-sawṭa ḍuriba bi-hi zaydun ('the whip [acc.], was Zayd hit with it?)and ʾāl-xiwāna ʾukila l-laḥmu ʿalay-hi ('the table [acc.], was the flesh eaten on it?'), 102alongside ʾa-zaydan summīta bi-hi ('Zayd [acc.], were you named so?').Whereas bi-in the latter is governed by the verb, 103 this is the case in neither the first nor the second sentence: in the first, the preposition bi-, conveying an instrumental meaning, is not governed by the verb ḍaraba; in the second, ʿalā conveys a locative meaning and is not governed by the verb ʾakala.Yet, Sībawayhi maintains that the prepositional phrase (or the genitive nominal alone) is fī mawḍiʿ naṣb, since if *ʾāl-sawṭa ḍuribta had been acceptable, the nominal introducing it would have been in the accusative, just like the pseudo-sentence *ʾa-zaydan marartu. 104) Although the unmarked choice for the constituent in an active sentence which is to function as the subject of the corresponding passive sentence is a direct object, e.g.ḍuriba zaydun ('Zayd was hit') (corresponding to e.g.[my illustration] ḍaraba ʿabdu llāhi zaydan ('ʿAbdallāh hit Zayd')), 105 grammarians also discuss constructions where it is a prepositional phrase, a ẓarf or al-mafʿūl al-muṭlaq, i.e. cognate accusative 106 (in the corresponding active sentence), which assumes this function (on conditions which will not be discussed here).The paradigmatic illustration for the first is sīra bi-zaydin ('Zayd was made to go'). 107owever, the following discussion by al-Mubarrad (al-Muqtaḍab, IV, 51-52) shows that this is not restricted to prepositional objects.He states, first, that when a preposition is attached to the mafʿūl, which precludes it from functioning as a subject, a ẓarf or a maṣdar may assume this function (which is impossible if the mafʿūl is an accusative nominal).Such is, for example, the construction sīra bi-zaydin sayrun šadīdun ('Zayd was made to go vehemently [nom.]')(or: yawmu l-jumʿati 'on Friday').Another illustration of this principle is ḍuriba bi-zaydin ʿišrūna sawṭan ('twenty [nom.]whip strokes [nom.] were struck due to Zayd'), whose prepositional phrase is explained as bi-sabab zayd or min ʾajli-hi.On the other hand, a few lines later, al-Mubarrad does allow the genitive nominal to function as the subject of a passive verb, notwithstanding its preposition, illustrating this with the sentence sīra bi-zaydin farsaxan ('Zayd was made to go a parasang').One may readily infer from this discussion that the possibility for such constituents to assume this function applies also to the preposition bi-zaydin in the sentence ḍuriba bi-zaydin, in which the preposition bi-, conveying the meaning of reason, is not governed by the verb ḍaraba. 108n a similar vein, Ibn Jinnī (al-Xaṣāʿiṣ, I, 397) adduces the following poetic verse as a case in which a prepositional phrase (viz.bi-ḏālika l-jirwi) 109 functions as the subject of a passive verb, although the sentence includes also a mafʿūl bi-hi ṣaḥīḥ, 110 viz.al-kilāba: wa-law waladat qufayratu jirwa kalbin la-subba bi-ḏālika l-jirwi l-kilābā.'had Qufayra given birth to a whelp, the dogs would have been reviled because of that whelp' This bi-, here also conveying the meaning of reason, is not governed by the verb sabba. 111inally, in a discussion of the preposition min conveying the meaning of tabʿīḍ ('indicating division into parts'), al-ʾAstarābāḏī (Šarḥ, IV, 265) proves his contention that in the construction ʾaxaḏtu mina l-darāhimi ('I took [some] of the dirhams'), from which the direct object (al-mafʿūl al-ṣarīḥ), e.g.šayʾan, is elided, 112 the preposition min 'depends' 113 on the verb (rather than functioning as an attribute to the direct object), by adducing the construction ʾuxiḏa mina l-darāhimi ('[some] of the dirhams were taken'), in which the prepositional phrase functions as the subject, 114 this notwithstanding the fact that this min cannot be regarded as governed by the verb. 115hrase, although the preposition might be elided, and in consequence the verb assigns the accusative to the (originally genitive) nominal, e.g.ixtāra (see §2.1); 116 and some are said to be free to take either an accusative nominal or a prepositional phrase, e.g.naṣaḥa, which can take either a direct object or a prepositional phrase introduced by li-(according to al-Zajjājī, al-Jumal, 31). 117owever, while one may readily hold that the preposition li-is governed by the verb naṣaḥa, al-Zajjājī (al-Jumal, 31) adduces two verbs which behave in the same fashion, yet, claiming that they govern the preposition li-seems to be rather forced; these are kāla and wazana. 118t thus follows that the difference between governed and non-governed prepositions has no bearing on the theory of ʿamal, the main pivot around which the whole Arabic grammatical tradition revolves. 119o recapitulate, these syntactic phenomena, which are common to both prepositional phrases and direct objects, are not restricted, with respect to the former, to prepositional objects.The linguistic facts do not furnish a solid syntactic foundation for distinguishing between governed and non-governed prepositions.
ʿabda llāhi120 qāʾiman ('ʿAbdallāh is in it [sc.the abode] standing') as a ẓarf.Al-ʾAstarābāḏī (Šarḥ, I, 243) explicitly refers to the grammarians' practice of labeling prepositional phrases as ẓarfs: commenting on Ibn al-Ḥājib's words wa-mā waqaʿa ẓarfan …, al-ʾAstarābāḏī says that Ibn al-Ḥājib means both ẓarfs and prepositional phrases, and that the reason for not mentioning the latter is that their syntactic behavior is just like the ẓarfs', to the extent that ẓarf is occasionally used as a term covering also prepositional phrases. 121he corollary from this practice, on the one hand, and from the discussion in previous subsections, on the other, is that prepositional phrases are entitled to be parsed both as mafʿūl bi-hi and as ẓarf, these two being, however, two terms denoting totally distinct functions.
To complicate things even further, grammarians frequently classify as ẓarfs prepositional phrases which do not convey any locative or temporal meaning.Ibn al-Sarrāj (al-ʾUṣūl, I, 205) explicitly treats as ẓarfs prepositional phrases which do not designate locations: wa-ʾiḏā kāna l-ẓarfu ġayra maḥallin li-l-ʾasmāʾi … ('and when the ẓarf does not denote a location of [the referents of] the nouns').This class is exemplified by the sentences fī-ka ʿabdu llāhi rāġibun ('you ʿAbdallāh desires'), min-ka ʾaxawā-ka hāribāni ('from you your two brothers are running away') and ʾilay-ka qawmu-ka qāṣidūna ('towards you your people are going'). 122The problem of 'double identity' pertains thus to all prepositional phrases.
We have also mentioned in §2.1 that whereas the term taʿaddin applies in Sībawayhi's al-Kitāb only to the relationship between a verb and a mafʿūl (bi-hi), in later treatises it acquires a double meaning: in the more restricted meaning it applies only to mafʿūl bi-hi, while in its more general meaning it applies also to constituents implementing other functions, e.g.ẓarf.If we consider the grammarians' conception of taʿaddin with regard to prepositional phrases, that is, that the basic function of prepositions is the taʿdiya of verbs to constituents with which they cannot engage in a taʿaddin relationship directly, then fī introducing locative or temporal prepositional phrases is no exception, and the term taʿaddin conveys the same meaning with regard to it as it does with regard to all other prepositions.This type of taʿaddin is on a par with the taʿaddin of verbs to direct objects, that is, it conveys the restricted significance.The inclusion of this fī as an ordinary member of the club of prepositions is conspicuous in the discussions grammarians dedicate to prepositions, where they refer to their functions and meanings (see §2.7).For instance, Ibn Jinnī in a chapter entitled bāb ḥurūf al-jarr (al-Lumaʿ, 30-31), after presenting a list of prepositions, including fī, asserts: fa-hāḏihi l-ḥurūfu tajurru mā tattaṣilu bi-hi wa-tuḍāfu ʾilay-hi ('these particles assign the genitive to what [sc. the noun which] they are connected with').This is followed by an illustrated list of prepositions and their meanings; the discussion of fī immediately follows min and ʾilā: wa-maʿnā fī l-wiʿāʾu wa-l-ẓarfiyyatu taqūlu zaydun fī ldāri wa-l-mālu fī l-kīsi ('the meaning of fī is a receptacle and of place/time qualification ["ẓarf-ness"; note that ẓarf also means 'receptacle'], e.g.zaydun fī l-dāri ['Zayd is in the abode'] and al-mālu fī l-kīsi ['the money is in the bag']'). 124That is to say, the syntactic behavior of fī is identical with that of other prepositions, and the meaning it conveys is deemed just another meaning of a preposition.On the other hand, the classification of these prepositional phrases as ẓarfs entails that the term taʿaddin, when applied to them, conveys the same meaning it does when applied to nominal ẓarfs, i.e. the general significance. 125oreover, grammarians occasionally put prepositional phrases conveying locative meaning, and consequently are to be parsed as ẓarfs, on a par, regarding their syntactic behavior, with prepositional phrases which are not so parsed.For instance, the occurrence of the preposition fī in prepositional phrases designating specific locations, e.g.al-masjid ('the mosque'), is explained by al-Jurjānī (al-Muqtaṣid, 643) in line with the two-stage model discussed above: since verbs which cannot engage in a taʿaddin relationship with nominals such as zayd are also unable to engage in a taʿaddin relationship with nominals such as almasjid, a preposition comes to the rescue, so instead of the ungrammatical *qaʿadtu lmasjida, one says qaʿadtu fī l-masjidi ('I sat in the mosque'), just as one says qaʿadtu ʾilā zaydin ('I sat with Zayd').In the same vein, al-Jurjānī (ibid, 274-275) explains that ẓarfs functioning as predicates are, in fact, clauses, due to the fact that the preposition fī (e.g. in fī l-dāri) needs some verb to 'depend' on (tataʿallaqu bi-hi, see §3.4) since the function of prepositions is to connect (li-tūṣila) verbs to nominals, as they do in the sentences qumtu ʾilā zaydin ('I stood up to be by Zayd') and ḏahabtu min dāri-ka ('I went out of your abode'). 126llustrations including locative or temporal prepositional phrases are also adduced in discussions pertinent to the four syntactic phenomena discussed in the previous subsection: (1) Conjunction constructions: After al-Jurjānī (ibid, 234-235) discusses the sentence marartu bi-zaydin wa-ʿamran, putting forward the two explanations mentioned above (while opting for the verb juztu, instead of ʾataytu), he adduces, for the same end, the following poetic verse: yaḏhabna fī najdin wa-ġawran ġāʾirā ('they travel in Najd [a high land] and in a low land [sc.Tihāma]'), accounting for the accusative of ġawr in the following manner: fa-ka-ʾanna-hu qāla yaslukna najdan wa-ġawran. 127The verb is regarded as al-muʿaddā bi-l-jārr, and the accusative is accounted for on the ground that the genitive nominal is mafʿūl with regard to al-maʿnā.This construction, it should be noted, is adduced in order to explain the ištiġāl construction, e.g.zaydan marartu bi-hi. 128) Ištiġāl constructions: In the chapter following Sībawayhi's (al-Kitāb, I, 32) discussion of ištiġāl constructions with prepositional phrases, entitled hāḏā bābu mā yajrī mimmā yakūnu ẓarfan hāḏā l-majrā ('this is a chapter on ẓarfs behaving likewise [i.e. in the same constructions discussed in the previous chapter]'), he discusses (ibid, I, 33-35), inter alia, ištiġāl constructions in which the preposition in question and the nominal introducing the sentence are ẓarfs.He states that it is permitted to say, for instance, yawma l-jumʿati ʾātī-ka fī-hi ('on Friday [acc.],I shall come to you then [lit."in it"]'), as it is permissible to say ʿabda llāhi marartu bi-hi.The nominal here is a ẓarf, and the underlying structure is *ʾalqāka yawma l-jumʿati [ʾātī-ka fī-hi] ('I shall encounter you on Friday, [I shall come to you then]').
In the same vein, after Ibn Hišām (Muġnī, 499) discusses ištiġāl constructions in which there is no other choice but to posit a verb different than the verb in the surface structure (e.g.zaydan marartu bi-hi, to which he posits the verb jāwaza), he asserts that when verbs may take either a direct object or a prepositional phrase, a verb identical with the verb in the surface structure may be posited, and hence the verb posited in the construction zaydan šakartu la-hu ('Zayd [acc.],I thanked him') is also šakartu: li-ʾanna šakara yataʿaddā bi-ljārri wa-bi-nafsi-hi ('since šakara engages in a taʿaddin relationship either through an operator or the genitive [sc. the preposition li-] or through itself').Immediately afterwards, he points out that this applies also to the ẓarf, as in the construction yawma l-jumʿati ṣumtu fīhi ('on Friday [acc.],I fasted then [lit."on it"]') (it is inferred that the underlying verb is also ṣumtu): li-ʾanna l-ʿāmila lā yataʿaddā ʾilā ḍamīri l-ẓarfi bi-nafsi-hi maʿa ʾanna-hu yataʿaddā ʾilā ẓāhiri-hi bi-nafsi-hi ('since the operator does not engage in a taʿaddin relationship with [lit."does not pass over to"] the pronouns of ẓarfs through itself, whereas it does so with overt ẓarfs'). 129) Prepositional phrases functioning as subjects of passive verbs: al-ʾAstarābāḏī (Šarḥ, I, 221) discusses the conditions which a ẓarf must fulfill in order to implement this function.The illustrative constructions adduced in his discussion include ʿind, fī l-dāri and fī-hā.130 (4) Verbs taking either a prepositional phrase or a direct object: al-Jurjānī (al-Muqtaṣid, 646) states that basically verbs of the category al-fiʿl ġayr al-mutaʿaddī engage in a taʿaddin relationship with nominals designating specific locations by means of a preposition, so that constructions lacking such a preposition are explained as cases of elision of the preposition.131 For instance, the underlying structure of the poetic verse … kamā ʿasala l-ṭarīqa l-ṯaʿlabu ('… as the fox ran on the way shaking') is … kamā ʿasala fī l-ṭarīqi [l-ṯaʿlabu], since al-ṭarīq designate a specific location.132 This problem of the double-identity of prepositional phrases reaches its peak in al-ʾIsfarāʾīnī (al-Lubāb, 84-85).This grammarian distinguishes between two types of al-mafʿūl bi-hi, direct and prepositional.The former separates al-mutaʿaddī from ġayr al-mutaʿaddī while on the latter he remarks: wa-yusammā ẓarfan ʾayḍan ('and it is also termed ẓarf').However, elsewhere (ibid, 81-83) he regards the ẓarf (= mafʿūl fī-hi), in line with grammatical tradition, as a separate category of al-manṣūb.133 Of the grammarians we examined, the only one who tackled this problem is al-ʾAstarābāḏī, who attributed to his predecessors a consistent theory with regard to the relationship of mafʿūl bi-hi, ẓarf and prepositional phrases, and propounded an alternative, extremely unorthodox, theory of his own on the matter.
A few lines after his assertion that the term ẓarf may apply also to prepositional phrases, in his discussion of the operator of ẓarfs (here only nominal ẓarfs are intended) and prepositional phrases functioning as predicates, he ascribes (Šarḥ, I, 244) to the Baṣrans the view that the ẓarf in these constructions takes the accusative as a mafʿūl fī-hi, just as it does in cases such as xarajtu yawma l-jumʿati ('I went out on Friday') (that is, where the verb is overt), whereas the prepositional phrase takes a virtual accusative as a 131 This is a case of ittisāʿ.132 See also Sībawayhi, al-Kitāb, I, 11; Ibn al-ʾAnbārī, ʾAsrār, 73-74; Ibn Hišām, Muġnī, 579, 637.A similar case is the problematic verb daxala (e.g. in the construction daxala fī l-dāri, see §2.5), which is occasionally discussed together with the also problematic ḏahaba l-šāma ('he went to Syria'), this latter construction being interpreted as stemming from the elision of ilā (see e.g.Sībawayhi, al-Kitāb, I,  11; Ibn al-Sarrāj, I, 171).However, grammarians do not parse fī l-dāri here explicitly as ẓarf, and occasionally even posit ilā as the elided preposition (see e.g.al-Jurjānī, al-Muqtaṣid, 600).Only two cases were found which could perhaps be taken as evidence that this prepositional phrase might be regarded (at least by the respective grammarians) as a ẓarf.First, al-ʾAstarābāḏī (Šarḥ, I, 492) asserts that daxala assigns the accusative to any nominal as a ẓarf, including in daxaltu l-dāra, which is explained as a case of elision of fī.One infers a fortiori that the underlying prepositional phrase is also a ẓarf.Incidentally, he attributes this view also to Sībawayhi (See also ibid, II, 369).The second instance is Ibn Hišām's (Muġnī, 159) discussion of a construction in which the verb wanā takes the preposition ʿan, said to convey in this case the meaning of ẓarfiyya: li-ʾanna wanā lā yataʿaddā ʾillā bifī ('since wanā does not engage in a taʿaddin relationship except through fī').The sentence wanā fī-hi is interpreted as daxala fī-hi wa-fatara ('he entered upon it but was remiss', wanā ʿan-hu meaning, for him, 'he passed from it, not entering upon it').However, in addition to the fact that this evidence is extremely convoluted, the term ẓarfiyya may be used here in a more general, non-technical, meaning.
(On Ibn Hišām's classification of ʿan as a preposition, probably in all cases except when preceded by min or ʿalā, see LEVIN 1987: 356-357; KASHER 2006: 169-171.)133 Incidentally, the same definition of al-mafʿūl bi-hi (with extremely similar wording) appears in al-Jurjānī, al-Taʿrīfāt, 241, yet this does not prevent the next entry in this book of definitions to be al-mafʿūl fī-hi (ibid, 242).
mafʿūl bi-hi, just as it does in cases such as marartu bi-zaydin (that is, here also, where the verb is overt).In other words, prepositional phrases (both locative/temporal and nonlocative/temporal, it is inferred) are, in fact, to be parsed as mafʿūl bi-hi, although they are occasionally referred to as ẓarfs, but only due to their syntactic behavior.His own theory is expressed elsewhere (ibid, I, 502-505).Here al-ʾAstarābāḏī discusses the grammarians' theory that the ẓarf may' become mafʿūl bi-hi, while keeping its meaning intact. 134This new status acquired by the ẓarf entails that its syntactic behavior is identical with a regular mafʿūl bi-hi.For example, when it undergoes pronominalization, it is not preceded by fī anymore, e.g.yawma l-jumʿati ṣumtu-hu ('on Friday [acc.],I fasted then [lit."it"]'). 135The climax of his discussion is reached when he states that the ẓarf 136 always possesses this status: the underlying structure of the sentence xarajtu yawma l-jumʿati is xarajtu fī yawmi l-jumʿati, in which the prepositional phrase is mafʿūl bi-hi by means of a preposition, so that after its elision, yawma l-jumʿati becomes mafʿūl bi-hi by means of no preposition, while keeping its (here: temporal) meaning intact.This is the same process of preposition elision which results in the construction istaġfartu llāha ḏanaban (see §3.3).The mafʿūl fī-hi (=ẓarf) is thus subsumed under mafʿūl bi-hi.According to this theory, the problem portrayed above, of whether one should parse prepositional phrases as mafʿūl bi-hi or as mafʿūl fī-hi, does not even arise.

Conclusion
The findings discussed in the present article show that what was termed here 'the standard categorization' of verbs according to their taʿaddin, ascribed to the Arab grammarians, is inaccurate: First, the term al-fiʿl al-mutaʿaddī bi-ḥarf jarr does not, in most cases, constitute a subcategory of al-fiʿl al-mutaʿaddī.Rather, it denotes the possibility of verbs to be connected, by means of prepositions, with constitutents with which they do not engage in a direct taʿaddin relationship.The preposition is regarded as an aid, enabling the verb to engage in such a taʿaddin relationship, as well as imposing a certain meaning on this relationship.
Second, the applicability of the term al-fiʿl al-mutaʿaddī bi-ḥarf jarr is not restricted to cases where the preposition at stake is governed.Rather, the basic function of all prepositions is regarded as 'transitivizing' verbs to constituents with which they do not engage in direct taʿaddin, irrespective of whether or not the preposition in question is governed by the verb; this notwithstanding the fact that grammarians were aware of the linguistic phenomenon of government of prepositions by verbs.
It was demonstrated how detrimental the lack of differentiation between governed and non-governed prepositions, with respect to the application of the terms taʿaddin and mafʿūl bi-hi, is to the notion of ẓarf.
The findings of the present article are based on grammatical treatises.A fuller picture of the notion of al-fiʿl al-mutaʿaddī bi-ḥarf jarr could perhaps be obtained by studying lexico-134 On this case of tawassuʿ, see fn. 9. 135 In contrast with yawma l-jumʿati ṣumtu fī-hi (an illustration which does not occur in this context).136 He maintains that this analysis pertains also to mafʿūl la-hu, i.e. the accusative of reason.
bi-] and the genitive nominal [sc.zaydin] together [sc. the prepositional phrase bi-zaydin] occupy the position of a [nominal in the] accusative [e.g.zaydan], due to the [syntactic effect of] the preceding verb [sc.marartu].') It should be noticed, first, what this excerpt does not say: it does not classify verbs into al-fiʿl al-mutaʿaddī bi-nafsi-hi vs. al-fiʿl al-mutaʿaddī bi-ḥarf jarr.As the heading ḏikr al-ʾasmāʾ al-manṣūbāt implies, Ibn al-Sarrāj's interest lies in accusative nominals, and accordingly, al-fiʿl al-mutaʿaddī refers only to verbs taking such constituents.This category stands in contradistinction to [al-fiʿl] llaḏī lā yataʿaddā.What is striking here is that the former is also referred to here by the phrase [fiʿl] yataʿaddā bi-ġayr ḥarf, whereas the latter is also referred to by the phrase [fiʿl] lā yataʿaddā ʾillā bi-ḥarf jarr.The verb qaʿada, for instance, belongs to the category allaḏī lā yataʿaddā (e.g.qaʿada ʿamrun); but it is also[fiʿl]lā yataʿaddā ʾillā bi-ḥarf jarr: if the speaker nevertheless wishes this verb to engage in a taʿaddin relationship with some nominal, he or she must use a preposition (e.g.qaʿada ʿamrun ʾilā bakrin), but this does not contradict the categorization of the verb as [al-fiʿl] allaḏī lā yataʿaddā.In other words, al-fiʿl al-mutaʿaddī bi-ḥarf jarr is not a category by itself, but is identical with al-fiʿl ġayr al-mutaʿaddī.Such categorization is in line with what was noted above regarding the 'negative' wording used bySībawayhi and al-Mubarrad: taʿaddin by means of a preposition does not constitute a positive trait of verbs, and consequently verbs are not categorized, with regard to their taʿaddin, according to the preposition(s) they take.In his more comprehensive book, Ibn al-Sarrāj (al-ʾUṣūl, I, 172) does not even mention this category when he subcategorizes al-ʾafʿāl al-mutaʿaddiya, here also in the chapter which deals with al-mafʿūl bi-hi (ibid, I, 169 ff.), under the heading ḏikr al-ʾasmāʾ al-manṣūbāt (ibid, I, 158).22However, when Ibn al-Sarrāj in this book deals with prepositions which are not otiose, 23 he states: [… li-ʾanna] l-ʾafʿāla llatī hiya ġayru mutaʿaddiyatin fī l-ʾaṣli lā tataʿaddā ʾillā bi-ḥarfi jarrin ('verbs which do not engage in a taʿaddin relationship in their basic state do not engage in a taʿaddin relationship except through a preposition') (ibid, II ʾaʿlama 'he made [someone] know'); (6) fiʿl lā yataʿaddā ʾillā bi-ḥarf xafḍ ('verb not engaging in a taʿaddin relationship except by means of a preposition') (e.g.daxala in daxaltu ʾilā ʾaxī-ka 'I entered upon your brother' 25 and marra in marartu bi-zaydin); (7) fiʿl yataʿaddā bi-ḥarf xafḍ wa-bi-ġayr ḥarf xafḍ ('verb engaging in a taʿaddin relationship either with or without a preposition') (e.g.naṣaḥa in naṣaḥtu zaydan and naṣaḥtu li-zaydin 'I advised Zayd').What is of interest here is that al-Zajjājī differentiates between categories (1) and (6), thereby recognizing a category of verbs taking prepositional phrases distinct from totally intransitive verbs; 26 recall that Ibn al-Sarrāj does not differentiate between [al-fiʿl] llaḏī lā yataʿaddā and [fiʿl] lā yataʿaddā ʾillā bi-ḥarf jarr, and consequently subsumes the verbs qāma and qaʿada under a single category which merges al-Zajjājī's categories (1) and (