ILḤĀQ AS A MORPHOLOGICAL TOOL
IN ARABIC GRAMMAR
The Arab grammarians differentiate between the ziyāda (augment) that introduces an element of meaning and the ziyāda that appends (yulḥiq) one morphological form to another. Having realized the potential of the concept of ilḥāq (appending) as an analytical tool in morphology, the grammarians divided appended words into several types according to the number of the radicals in their roots and the type of ziyāda that is involved, and tried to justify forms and patterns with reference to a set of detailed rules which they elaborately describe. This paper deals with the issues the grammarians tackle in their study of ilḥāq, such as its purpose, the possibility of analogically extending its examples, and the inapplicability of idġām (gemination) to its patterns. It also examines how the grammarians use ilḥāq to reduce considerably the number of morphological patterns that form a closed system, to explain away anomalous and rare patterns, and thus to limit deviation from the norm (qiyās) and to test the validity of a host of morphological issues.
1.1. Within the Arabic root system a consonant may either be a radical (aṣl) or an augment (ziyāda), i.e., part of the etymological root or some kind of morphological affix, respectively. In discussing augmented forms, grammarians usually differentiate between the purely morphological ziyāda, whose purpose is to introduce an element of meaning, and the ziyāda whose purpose is to append (yulḥiq) one formal word pattern to another by interpreting one or more consonants in the word as having the status of affixes and not radicals. As with many postulates, this distinction goes back to Sībawayhi (d. 180/796; Kitāb, II, 9), and probably to his teacher, al-Ḫalīl b. Aḥmad (d. 175/791), whose influence on him was overwhelmingly in the areas of phonology and morphology (cf. Carter 1973, 154, and 1981, 352). To clarify this distinction, the grammarians had not only to define the limits between the two types of ziyāda in view of both form and meaning, but also to justify why certain words could not be considered appended (mulḥaq), although their forms do suggest such a possibility.
This painstaking task which the grammarians shouldered, and which  necessitated close scrutiny of a host of mostly complex and rarely used words whose patterns are said to be the result of ilḥāq (appending), was motivated by their general tendency towards limiting the items that constitute a closed system—particularly, the number of patterns the available corpus of words should be divided into—and by their interest in using the rules that govern ilḥāq as testing devices to prove the validity of their more general morphological premises. This paper sets out to investigate the methods the grammarians used in their study of ilḥāq and to demonstrate how they tried to incorporate these rules within their overall system of morphological analysis.
1.2. Much of the material on ilḥāq is discussed in several scattered parts of the Kitāb (esp. II, 8–11; 197; 334–41; 401–403), but Sībawayhi nowhere gives a definition of ilḥāq or formulates and lists together the rules that pertain to it. Equally scattered are the comments of Mubarrad (d. 285/898) in his Muqtaḍab (esp. I, 204–205, 244; II, 225ff; III, 88, 385–86; IV, 3–4). Māzinī (d. 248/862), on the other hand, discusses the different aspects of ilḥāq in one part of his Taṣrīf (I, 34–53), be it in less detail than in the Kitāb or the Muqtaḍab. However, Ibn Ǧinnī’s (d. 392/1002) commentary on the Taṣrīf complements its text to make it more or less comprehensive. Furthermore, as we shall see later, Ibn Ǧinnī makes several perceptive observations on ilḥāq as part of his unparalleled approach to linguistic analysis.
As for the most well-organized and comprehensive study of ilḥāq in the sources, it is obviously that of Astarābāḏī (d. 686/1287) in Šarḥ al-Šāfiya (I, 52–70). It is surprising, however, that some authors of major works on morphology barely mention a few rules about ilḥāq, as did Ibn ʿUṣfūr (d. 669/1271) in his Mumtiʿ (I, 206–208), or sporadically mention its function without devoting a particular section or chapter to it, as did Ibn Ǧinnī, who at times mentions, in his alphabetical list of ḥurūf (here, phonemes) in Sirr ṣināʿat al-iʿrāb, that a certain ḥarf can have the function of ilḥāq (e.g., alif; II, 691–93).
 Many of the later sources also show little interest in ilḥāq, and it is remarkable that, unlike his commentary on Ibn Ǧinnī’s al-Taṣrīf al-mulūkī (64f.), Ibn Yaʿīš’s (d. 643/1245) most detailed work, Šarḥ al-Mufaṣṣal, does not include a special chapter on ilḥāq (see sporadic mention of the term in VI, 113, 119, and IX, 146–48; cf. Zamaḫšarī, Mufaṣṣal, 240, 241, and 358, where the term ilḥāq appears only in the latter case). Also noteworthy is that Suyūṭī (d. 911/1505) has an atypically short summary of the main issues of ilḥāq in his Hamʿ al-hawāmiʿ (II, 216–17)—most of which relates to whether or not it is restricted to what the Arabs actually used—and only an incidental mention of ilḥāq as one of the kinds of ziyāda in Ašbāh (IV, 137).
Finally it should be mentioned that works that deal with loan words usually cite ilḥāq as one of the main factors that affect the Arabicized forms of these loans. For example, Ǧawālīqī (d. 540/1145) and Ḫafāǧī (d. 1069/1659) mention several patterns that demonstrate this phenomenon (Muʿarrab, 8, and Šifāʾ, 36–37), and Ibn Kamāl Pasha (d. 940/1533) has a lively discussion of its role in Arabicization and frequently refers to this role in analyzing particular examples (Risāla, 47f.; and index, p. 153). This interest in the relation between ilḥāq and loan words, it may be noted, owes its origin to Sībawayhi’s chapter on mā uʿriba min al-aʿǧamiyya (What has been Arabicized from foreign languages; Kitāb, II, 342).
Since the above-mentioned authors are largely in agreement concerning the function of ilḥāq and the material that constitutes its corpus, we shall refer to them collectively unless we need to specify or indicate different views.
2.1. Although Sībawayhi does not give a formal definition of ilḥāq, his discussion of it includes all the elements later grammarians used in formulating its definition. These elements are the following: (a) that it is a ziyāda; (b) that it causes triliterals to be appended to quadriliterals and quinqueliterals, and quadriliterals to be appended to quinqueliterals; (c) that this ziyāda is different from the one which uniformly introduces an element of meaning; (d) that the pattern of the appended word should  phonologically conform to the pattern of the word to which it is appended, i.e., what can be referred to as the target pattern; (e) that the derivatives of the appended word should be congruent to the derivatives of the target word; and (f) that the rules of assimilation (idġām), if applicable, should not be made operational in the appended word because this would change its pattern and hence its congruence to the word to which it is appended. Due to the highly complex nature of the subject, sections 2.2–2.4 will deal in more detail with the grammarians’ views on the above elements, and we shall try later to examine issues of a more general nature that relate to the grammarians’ use of this tool in their linguistic analysis.
2.2. The phonemes used for appending are mostly wāw, and yāʾ (e.g., kawṯar and ḍayġam, both appended to ǧaʿfar; and ḫirwaʿ, and ʿiṯyar, both appended to dirham), but they also include, among others, nūn (e.g., raʿšan, appended to ǧaʿfar), mīm (e.g., dilqim appended to zibriǧ), and alif (e.g., ḥabanṭā—from the root ḤBṬ, with the addition of nūn and alif—which is appended to ḥabarkā, itself with a final alif that is not part of the root). Such instances of augmentation with no recurring phonemes are often attributed to their basic roots by a semantic comparison between their apparent root and an assumed root with less radicals. A good example is that of dulāmiṣ (shining), whose apparent root DLMṣ is further reduced to a triliteral root semantically related to it, DLṣ (cf. dalīṣ, daliṣ, dilāṣ, and dalāṣ, all of which mean “shining”; see Ibn Manẓūr, Lisān, DLṣ, and Zubaydī, Amṯila, 62). Thus, the mīm, according to the grammarians, appends dulāmiṣ to ǧuḫādib, an authentic quadriliteral.
In addition to this, theoretically any phoneme can be used for appending if it recurs within the appended word. Examples include mahdad, ḫidabb, ʿaṯawṯal, ḥalakūk, qurṭāṭ, ʿafanǧaǧ, and qušaʿrīra appended to ǧaʿfar, qimaṭr, farazdaq, qarabūs, qurṭās, ʿabanqas, and ḫuzaʿbīla, respectively (Suyūṭī, Muzhir, II, 35–36). The difference between  the first seven representative examples and their respective counterparts is that all members of the first group may be attributed to roots whose radicals are supposedly reducible to a number less than the number of radicals in the root of their counterparts to which they are appended. For example, ʿaṯawṯal (stout, fleshy, and flabby) is apparently quinqueliteral like farazdaq, but since it is semantically related to the root ʿṮL (which indicates abundance, stoutness, flabbiness, etc., and which was augmented by the addition of wāw and ṯāʾ, according to the pattern faʿallal), it is considered triliteral in origin, unlike the loan word farazdaq, none of whose radicals may be reducible with reference to a triliteral or quadriliteral root to which it may be assigned. Similarly, ḥalakūk (intensely black) is derived from a triliteral root ḤLK which indicates blackness, whereas its counterpart, qarabūs (part of a horse’s saddle), also a loan word, is thought to have four radicals (q, r, b, and s) that must be considered part of its supposed root.
Based on the above, the vast majority of the corpus of appended words may be divided into five types: 
i. Triliterals appended to quadriliterals: e.g., ǧadwal (ǦDL) and raʿšan (RʿŠ), compared with ǧaʿfar (ǦʿFR); and duḫlul (DḪL) and ḥulkum (ḤLK), compared with burṯun (BRṮN).
ii. Triliterals appended to augmented quadriliterals: e.g., dulāmiṣ (DLṣ), compared with ǧuḫādib (ǦḪDB); ʾiḥlīl (ḤLL), compared with birṭīl (BRṬL); and ḥabawnan (ḤBN), compared with ḥabawkar (ḤBKR).
iii. Triliterals appended to quinqueliterals: e.g., ʾinqaḥl (QḤL), compared with qirṭaʿb (QRṬʿB); and ḥabarbar (ḤBR), compared with farazdaq (FRZDQ).
iv. Quadriliterals appended to quinqueliterals: e.g., qiršabb (QRŠB), compared with qirṭaʿb (QRṬʿB); and ǧaḥanfal (ǦḤFL), compared with safarǧal (SFRǦL).
v. Quadriliterals appended to augmented quinqueliterals: e.g., qušaʿrīra (QŠʿR), compared with ḫuzaʿbīla (ḪZʿBL); and ḫaysafūǧ (ḪSFǦ), compared with ʿaḍrafūṭ (ʿḌRFṬ).
Since words cannot have more than five radicals, ilḥāq does not affect quinqueliterals (li-anna banāt al-ḫamsa laysa warāʾahā šayʾ min al-aṣl fa-yulḥaq bi-hi; Ibn Ǧinnī, Munṣif, I, 51). In other words, because there is no target pattern which the quinqueliterals can be appended to, ilḥāq was not applied to them, and they had to be placed outside the closed system which ilḥāq represents (see 3.3 below).
2.3. At the level of meaning, the grammarians draw a sharp distinction between ilḥāq and augmentation through which patterns that indicate certain meanings are formed. Of course, this latter type is much more widespread than ilḥāq and may be viewed as derivation (ištiqāq) par excellence, whereas ilḥāq is a special type of derivation whose relative frequency of use is quite limited. This not withstanding, the grammarians consider the two types to be on an equal footing in the process of deriving words since they consider each of them to be representative of a distinct purpose of ziyāda. As noted in 1.1 above, Sībawayhi (Kitāb, II, 9)  alluded to the distinction between two kinds of ziyāda, one of which appends one form to another (tulḥiq bināʾan bi-bināʾ), while the other introduces an element of meaning (tadḫul li-maʿnā).
Māzinī also makes this distinction (Taṣrīf, I, 13, and Ibn Ǧinnī, Munṣif, I, 13–17), but in a less direct way. Based on the purpose of ziyāda, he classifies it into four types: (a) the ziyāda of ilḥāq, which appends one form to another; (b) the ziyāda for vowel prolongation, such as ʿaǧūz and ǧarīb; (c) the ziyāda that indicates a meaning (maʿnā), such as nūnation (tanwīn) and the prefixes of the imperfect (ḥurūf al-muḍāraʿa); and (d) the ziyāda that is inseparable from the word because the very meaning (maʿnā) of the word is dependent on the augmented pattern, e.g., the alif and tāʾ of iftaqara, which have been part of the pattern iftaʿala since it was first coined (wuḍiʿa) and used instead of *faqura. A closer look at this apparently more elaborate classification, however, readily reveals that it is essentially consistent with Sībawayhi’s, since it contrasts ilḥāq with the ziyāda that indicates meaning. Of the latter type are (c) and (d) above, where the word “meaning” is used in Māzinī’s text, as well as (b), since vowel prolongation is part of the structure of several patterns that are indicative of meaning, as in Māzinī’s own example, ʿaǧūz, of the pattern faʿūl, which indicates a common adjective for both masculine and feminine, and which has a plural, ʿaǧāʾiz, that is exclusively indicative of the feminine (Sībawayhi, Kitāb, II, 208; cf. II, 131, where the wāw in ʿaǧūz is contrasted with the ziyāda of ilḥāq).
Ibn ʿUṣfūr’s classification of the types of ziyāda (Mumtiʿ, I, 204–6) is even more elaborate than Māzinī’s, since it includes types that are either purely phonological, such as the hāʾ of quiescence (hāʾ al-sakt), or that do not strictly qualify for inclusion under separate headings, such as the feminine ending of zanādiqa—called tāʾ (or hāʾ) of compensation (tāʾ al-ʿiwaḍ) on the assumption that it compensates for the elided yāʾ in zanādīq—which actually belongs to a pattern that indicates the plural, and hence meaning. Taking this into consideration, the core of Ibn ʿUṣfūr’s classification is basically in agreement with that of Sībawayhi and Māzinī.
The distinction of the grammarians between the ziyāda of ilḥāq and the ziyāda of maʿnā raises the problem of those appended words which  apparently do carry an element of meaning due to their augment. An example of such words is ḥawqala (said of a man who ages and becomes weak), which is appended to fawʿala (Ibn ʿUṣfūr, Mumtiʿ, I, 167), and whose meaning is not identical to the original verb, ḥaqila (said of a camel that suffers indigestion after drinking water mixed with sand). To resolve this discrepancy, Astarābāḏī uses this example, among others (Šarḥ, I, 52–53), to introduce a vital component to the definition of ilḥāq, and hence to the distinction between ilḥāq as a ziyāda that is described as not having to do with meaning and the ziyāda that indicates meaning. In his terms, the ziyāda of ilḥāq is ġayr muṭṭarida fī ifādat maʿnā, that is, it does not systematically add a well-defined element of meaning. It is this unsystematic characteristic of ilḥāq that truly distinguishes it from the ziyāda that systematically introduces a discernible element of meaning and is therefore outside the sphere of ilḥāq, as is the case in the hamza of ʾakbar and ʾafḍal, which, he says, consistently expresses the comparative (tafḍīl), and the mīm of the pattern mifʿal, which consistently indicates the instrument (Šarḥ, I, 53; II, 332).
2.4. At the purely formal (lafẓī) level, the grammarians identify several rules associated with ilḥāq. These rules, scattered as they are in the earlier sources, were assembled by some later authors either to formulate an accurate definition of ilḥāq, as did Astarābāḏī (Šarḥ, I, 52), or to list each criterion (ḍābiṭ) that reveals the use of ilḥāq, as did Suyūṭī (Hamʿ, II, 216). In this respect too, the grammarians seem to be most interested in the distinction between ziyāda of ilḥāq and ziyāda of maʿnā. Indeed, their discussion centers on two main aspects which endorse their distinction of the two types. The first aspect is the congruence between the appended word and the word to which it is appended with regard to the number of radicals and the metric measure (wazn), i.e., the pattern of ḥarakāt and sakanāt (occurrence or non-occurrence of vowels after consonants). This congruence, the grammarians stress, should also apply to the derivatives of both words, that is, in the case of verbs (usually cited in the perfect), it should manifest itself in the imperfect, the imperative, the verbal noun, the active participle, and the passive participle, and in the case of nouns, it should appear in the diminutive and broken plural forms. Without going into details and exceptions to this general guideline of congruence, suffice it here to say that it was used to show the underlying difference between what is mulḥaq and what is not.
For example, Astarābāḏī (Šarḥ, I, 55; cf. Ibn Ǧinnī, Ḫaṣāʾiṣ, I, 222, 232) argues that the inclusion of the verbal noun in the above list of derivatives  that manifest congruence between the words that are appended and the words they are appended to should disqualify patterns such as afʿala, faʿʿala, and fāʿala from being appended to the verb daḥraǧa, with which they rhyme, since their verbal nouns, ifʿāl, tafʿīl and mufāʿala, are not congruent with the verbal noun of daḥraǧa, which is daḥraǧa(tun), of the pattern faʿlala(tun). Similarly, in nouns, the insistence that congruence should apply in broken plurals, according to Astarābāḏī (ibid., I, 56), readily shows that ḥimār, in spite of being metrically equivalent to qimaṭr, cannot be considered appended to it, since its broken plurals, ḥumur and aḥmira, are not of the same pattern as qamāṭir. Obviously, the inclusion of such peculiarities of ilḥāq in its definition in Astarābāḏī’s Šarḥ (I, 52) is the reason why this definition, whose aim is to exclude other phenomena, is unusually long and detailed.
The second aspect which the grammarians focus on in the distinction between the ziyāda of ilḥāq and the ziyāda of maʿnā at the formal level is that idġām (gemination) applies to the latter but not to the former. Sībawayhi notes the difference between these two types of ziyāda as to the applicability of idġām, and devotes a chapter to those appended words whose final radicals are reduplicated but not geminated (Kitāb, II, 401–402; cf. II, 408). Thus, he contrasts qardad, which is appended to ǧaʿfar and salhab, with maradd, originally *mardad, and attributes the lack of idġām to ilḥāq itself. The aim of the contrast between qardad and *mardad is to show that idġām, for which both words qualify according to their phonological structure, becomes inoperable in the presence of ilḥāq. In fact, Sībawayhi argues that idġām does not take place in such appended words because the speaker intentionally keeps the last two radicals separate in order to achieve ilḥāq through the augment (lam tudġim li-annaka innamā aradta an tuḍāʿif li-tulḥiqahu bi-mā zidta bi-daḥraǧtu wa-ǧaḥdaltu). This is why in ǧalbaba, he says, the two bāʾs are not geminated, hence the use of the forms ǧalbabtuhu, muǧalbab, ǧulbiba, taǧalbaba, yataǧalbabu, and the like, which are appended to their counterparts derived from daḥraǧa, such as tadaḥraǧa, yatadaḥraǧu, and daḥraǧtu (Kitāb, II, 401; cf. Fārisī, Taʿlīqa, V, 156–57; Mubarrad, Muqtaḍab, I, 204–205, 244). Ibn ʿUṣfūr (Mumtiʿ, I, 207)  lends further support to this line of thinking by arguing that speakers tolerated the heaviness of two separate radicals (iḥtamalū ṯiqal iǧtimāʿ al-miṯlayn) in such examples in order for their patterns to remain congruent to the words to which they were appended.
This structural identity, so to speak, of appended words was viewed by Ibn Ǧinnī—whose unremitting quest for exploring the underlying principles of linguistic phenomena is largely unrivalled in the Arabic tradition—as part of a more general tendency which he detects in a host of examples (Ḫaṣāʾiṣ, III, 232–40; esp. 232–33) and which he discusses under the title al-imtināʿ min naqḍ al-ġaraḍ (refusal to contradict the objective). The essence of his argument, in the case of ilḥāq, is that its objective of achieving congruence between appended words and what they are appended to would have been annulled if normal idġām been applied, and thus the Arabs refrained from applying the rules of idġām to appended words because it was necessary to protect (ḥirāsa) and preserve (ḥifẓ) the original purpose. Apart from the fact that this explanation presupposes a conscious effort on the part of the speaker, its inclusion with allegedly comparable phenomena is an attempt to show that ilḥāq, which represents an anomalous case with regard to the rules of idġām, is not necessarily anomalous in other respects. In connection with this, we shall try to show later (see 4.2 below) how the grammarians incorporated the phenomenon of ilḥāq within the general grammatical system, as they saw it, by applying to it the same criteria of analysis that they use in other cases.
3.1. Based on the elements that they included in defining ilḥāq (see 2.1 above), and on their distinction between the ziyāda of ilḥāq and the ziyāda of maʿnā both at the level of meaning and form (2.3 and 2.4 respectively), the later grammarians were well-disposed toward assigning to ilḥāq an ultimate purpose that would justify its existence as an independent phenomenon. In this respect, it seems that they wanted to surpass the earlier grammarians, who merely stated that the ziyāda of ilḥāq appends one word to another (tulḥiq bināʾan bi-bināʾ; see Sībawayhi, Kitāb, II, 9, and Māzinī, Taṣrīf, I, 13) and did not go beyond this self-explanatory level to determine a more specific purpose for ilḥāq. The usual view among the later grammarians is that the ultimate purpose of  this ziyāda is to accommodate the use of the language, particularly in rhymed prose (saǧʿ) and poetry, with ittisāʿ or tawassuʿ (lit., latitude of speech). This view, which is attributed by Ibn Ǧinnī to his teacher Abū ʿAlī al-Fārisī (d. 377/987) seems to have been generally, but not universally, adopted in the sources (cf. Ibn Ǧinnī, Ḫaṣāʾiṣ, I, 358; II, 25, and Munṣif, I, 34, 38, 43; Ibn Yaʿīš, Šarḥ al-Mulūkī, 65; Astarābāḏī, Šarḥ, I, 66–67; Suyūṭī, Hamʿ, II, 217). The problem with this interpretation, however, is twofold. Firstly, the proposed ittisāʿ could only be achieved if the writer or poet were free to apply analogy and come up with words that may never have been heard before. Fārisī was aware of this prerequisite and tried to circumvent it by asserting that one may invent such words, on the analogy of attested examples, and thus use, in poetry, constructions like ḍarbaba Zaydun ʿAmran, marartu bi-raǧulin ḍarbabin, and ḍarbabun afḍalu min ḫarǧaǧin, where ḍarbab is used as a verb, an adjective, and a noun, respectively (Ibn Ǧinnī, Munṣif, I, 43–44; cf. Ḫaṣāʾiṣ, I, 358–59).
The inadequacy of this argument is nonetheless evident to Ibn Ǧinnī, who alludes to his teacher’s view, both in Ḫaṣāʾiṣ and Munṣif, as part of his discussion of Māzinī’s distinction between those appended forms that are qiyāsī (regular, analogically extended) and those that are samāʿī (unproductive, restricted to attested material). In fact, Ibn Ǧinnī seems to alert the reader to the limited applicability of Fārisī’s view. He does this not only by giving an account of their discussion, during which Ibn Ǧinnī asks whether it would not be tantamount to inventing speech (a-fatartaǧil al-luġa irtiǧālan; Ḫaṣāʾiṣ, I, 359; cf. Munṣif, I, 44), but also by supporting Māzinī’s view that analogical extension does not apply to any of the appended forms other than those of the pattern faʿlal, such as mahdad and ǧalbab, where the third radical is duplicated (Munṣif, I, 42), and thus forms like ǧawhar, bayṭar, ǧadwal, ḥiḏyam, rahwak, arṭā, miʿzā, salqā, and ǧaʿbā (Ḫaṣāʾiṣ, I, 358) are restricted to samāʿ. Moreover, the issue of the qiyāsī versus the samāʿī nature of ilḥāq is presented by Suyūṭī (Hamʿ, II, 217; cf. Ibn Mālik, Tashīl, 299) as a subject of controversy among three parties. The first of these restricts ilḥāq to samāʿ, unless the grammarians need to create words with which to train students, whereas the second party—to which Fārisī belongs—puts no restraints on analogically extending its attested examples. The third party is more selective since it resorts to the criterion of frequency of usage to determine the permissibility, or otherwise, of allowing analogical extension.
The other problem related to Fārisī’s view that ittisāʿ is the ultimate  purpose of ilḥāq is that, as we learn from some grammarians, the same notion of ittisāʿ can explain other types of ziyāda and, conversely, that some specimens of ilḥāq are explicable by alternative notions used to explain other types of ziyāda. The first part of this problem is most clearly visible in Ibn al-Ḥāǧib’s (d. 646/1249) text (in Astarābāḏī’s Šarḥ, I, 65–66), where he assigns different purposes for the existence of patterns, including augmented ones, and cites tawassuʿ as the reason for using the maqṣūr (abbreviated) and mamdūd (prolonged) forms (i.e., in certain doublets), as well as what he calls ḏū l-ziyāda (augmented [word]). The other part of the problem is evident in Astarābāḏī’s explanation of Ibn al-Ḥāǧib’s text, since he asserts that the notion of ḥāǧa (need), rather than tawassuʿ, is the real purpose of using the ziyāda of ilḥāq as well as other kinds of ziyāda, such as that of the active participle, the passive participle, and the verbal noun. Astarābāḏī then hastens to say that it is also possible to explain the ziyāda of ilḥāq by tawassuʿ.
It is evident from the above that the grammarians, starting with Fārisī, were trying to justify the existence of ilḥāq by assigning a purpose to it, just like other morphological phenomena which they associated with distinctive purposes. By insisting, however, on determining this purpose more specifically than did earlier grammarians such as Sībawayhi and Māzinī, who merely stated that ilḥāq appends certain words to others, these grammarians actually failed to appreciate why their forerunners acknowledged ilḥāq as a distinct phenomenon, and consequently why they contented themselves, in determining its purpose, with explaining what it does, and stopped short of seeking a more specific purpose to ascribe to it.
3.2. In order to understand the significance of ilḥāq for the earlier grammarians, and particularly with regard to the difference we have just mentioned between them and their successors, it is more appropriate to speak of the role that they assigned to ilḥāq in their analysis than of the purpose that it serves from the angle of the speaker. In other words, the early grammarians, most notably Sībawayhi and Māzinī, treated ilḥāq as a phenomenon in its own right and did not consider it to be part of any larger phenomenon because they realized its huge potential as a tool of morphological analysis. Theoretically speaking, they could have considered it to be a kind of ištiqāq whose examples are characterized by the  use of certain phonemes and/or the repetition of others, and so on (e.g., say that ḥalakūk is derived from the root ḤLK with the introduction of the long vowel ū and the repetition of k). Alternatively, they could have said, as some lexicographers did (see n. 6), that these examples are of the same patterns as those words to which they are appended and could thus effectively have avoided the need to speak of ilḥāq (e.g., ḥalakūk would be—regardless of the number of its radicals as explained in 2.2 above—on an equal footing with qarabūs, both of the pattern faʿalūl, and not appended to it). The fact that they chose to think of it as a distinct phenomenon, therefore, was not dictated by the nature of the corpus of words that were considered to be examples of it—unlike, for example, the three other kinds of ziyāda that Māzinī mentions (see 2.3 above) and that are linguistic realities that naturally represent undeniable and self-explanatory distinct phenomena. In effect the early grammarians were responding to their own interest in what they perceived as a major analytical tool. This explains why they were not concerned with what its purpose is from the point of view of the speaker. As for the later grammarians’ search for a purpose for ilḥāq grounded in pragmatics rather than pure analysis, it is now evident that it went against the very reason why the earlier grammarians recognized ilḥāq as a distinct phenomenon.
The most obvious advantage that ilḥāq represented for the earlier grammarians is that it enabled them to reduce considerably the number of what we can describe as major morphological patterns that they had to acknowledge within a closed system. A quick look at the list of words that are said, in any grammatical work that includes them, to be appended to the word that represents such a pattern readily reveals the extent of this reduction. In the case of the major pattern faʿlal, for example, Ibn al-Sarrāǧ (Uṣūl, III, 182) gives ǧaʿfar and salhab as the noun and adjective that represent it and to which other words are appended. These words, the supposed radicals of whose roots are considered to be reducible to less than the four radicals of ǧaʿfar and salhab (see 2.2 above), are: ḥawqal (fawʿal), zaynab (fayʿal), ǧadwal (faʿwal), mahdad (faʿlal), ʿalqā (faʿlā), raʿšan (faʿlan), sanbata (faʿlat or fanʿal), and  ʿansal (fanʿal). The eight different patterns which these eight words represent were thus grouped together under one major phonological pattern, faʿlal, since all of them conform to its wazn (metric measure), that is, its pattern of ḥarakāt and sakanāt (see 2.4 above). Similarly, hundreds of words are then cited by Ibn al-Sarrāǧ and grouped in such major patterns (ibid., III, 181–222). Further reduction in the number of patterns is achieved by the grammarians’ acknowledgement of the possibility of appending triliterals to quadriliterals that are, themselves, appended to quinqueliterals (Kitāb, II, 341)—such as ʿafanǧaǧ (root ʿFǦ) which is appended to ǧaḥanfal (root ǦḤFL), itself appended to safarǧal (root SFRǦL)—and of deriving appended words from other appended words (Astarābāḏī, Šārḥ, I, 55)—as tašayṭana, which is appended to tadaḥraǧa and is derived from šayṭana, itself appended to daḥraǧa. Understandably, the grammarians halted the process of reduction with words that are augmented quinqueliterals, such as qabaʿṯarā, simply because they did not find a six-radical pattern to which they could append them, and so there was no possibility of grouping words under major patterns (cf. Sībawayhi, Kitāb, II, 9; Fārisī, Baġdādiyyāt, 122, 434; Ibn Ǧinnī, Munṣif, I, 51, and Ḫaṣāʾiṣ, I, 319–20; Ibn ʿUṣfūr, Mumtiʿ, I, 206).
The grouping of appended words into major patterns gave the grammarians another considerable analytical advantage, namely, that they were able to draw up rules that are applicable not only to the words that represent the pattern and are not themselves appended to other words—e.g., ǧaʿfar and salhab of the major pattern faʿlal mentioned in the previous paragraph—but also to all the words whose patterns are appended to faʿlal—e.g., ḥawqal, ʿawsaǧ, zawraq, hawdaǧ, etc., which are of the pattern fawʿal, and zaynab, ġaylam, ṣayraf, ḍayġam, etc., which are of the pattern fayʿal, and so on. Such rules are abundant in the sources, as  early as Sībawayhi’s Kitāb. For example, Sībawayhi formulates a universal rule to the effect that all triliterals that were augmented to become quadriliterals and were appended to genuine quadriliterals have, like these quadriliterals, broken plurals of the pattern mafāʿil, such as ǧadwal, ʿiṯyar, kawkab, tawlab, sullam, dummal, ǧundab, and qardad, whose plurals are ǧadāwil, ʿaṯāyir, kawākib, tawālib, salālim, damāmil, ǧanādib, and qarādid, respectively (II, 197, and Fārisī, Taʿlīqa, IV, 95; see other examples in ʿUḍayma’s Fahāris, 364–72). The ultimate application of such rules may be seen in the pattern lists that some sources have (e.g., Ibn al-Sarrāǧ, Uṣūl, III, 181f., and Suyūṭī, Muzhir, II, 6f.). In such lists, the grammarians group together appended words with the words that they are appended to and present each group as a homogenous category that shares several morphological traits applicable to all its constituents, irrespective of whether they are appended words or not.
3.3. The grammarians’ use of ilḥāq as an analytical tool shows that they also employed it to achieve one of their principal goals—to limit deviations from the norm (qiyās) and maximize the applicability of grammatical rules. The necessary condition for this purpose to be achieved, in the case of ilḥāq, is the existence of a closed and well-defined system that would unmistakably identify appended words and patterns and describe the rules to which they are subject. Once this is accomplished, words that do not conform to these rules can be easily disqualified from inclusion in the closed system.
The mere fact that ilḥāq involves the condensation of several patterns into one major pattern (see 3.2 above) goes a long way towards reducing the examples to a more manageable number. Moreover, a sizeable portion of the corpus of appended words represents extremely rare usages which, after being appended to major patterns, become effectively part of the norm of their own class and, consequently, cease to stand out as extremely rare or solitary examples, as they indeed were prior to the classification process of ilḥāq. One such example is hammariš (adjective for a very old and wrinkled woman; e.g., ʿaǧūz hammariš), which represents  a pattern faʿʿalil described by Sībawayhi as qalīl (Kitāb, II, 339) and which may be the only quadriliteral example of this pattern (Ibn Manẓūr, Lisān, HMRŠ). Once this pattern is appended to a quinqueliteral word, such as qahbalis, ǧaḥmariš, and ṣahṣaliq (Kitāb, II, 341, 354; Suyūṭī, Muzhir, II, 35), it becomes part of a larger entity and is no more regarded as anomalous. The same can be said of other examples usually cited as appended words, such as naḫwariš, bulahniya, firindād, ḫayzalā, ʿilwadd, and others.
Other techniques the grammarians employed in matters related to ilḥāq should also be seen in the light of their effort to limit deviation from the norm. Sībawayhi’s treatment of qīqāʾ and zīzāʾ involves one such technique. Now these two words belong to the category of ism (noun), as opposed to maṣdar (verbal noun), and so the word to which they are to be appended should also be an ism, in line with the regular distinction Sībawayhi—and the later grammarians—drew between ism and maṣdar in their study of ilḥāq. The anomaly in the case of zīzāʾ and qīqāʾ, however, is that the pattern to which they should be appended—the reduplicated biliteral of fiʿlāl, i.e., *fiʿfāʿ, such as qilqāl—is used exclusively with maṣdars (Sībawayhi, Kitāb, II, 386; Māzinī, Taṣrīf, II, 180; Ibn ʿUṣfūr, Mumtiʿ, I, 151). To avoid this anomaly, which would affect the applicability of the distinction between ism and maṣdar, Sībawayhi appends these two words to the nearest hamzated and unduplicated fiʿlāl pattern (i.e., fiʿlāʾ) that does occur with isms, and chooses ʿilbāʾ to illustrate it. Another technique that ensures the widest possible application of qiyās is the analogical extension of the rule (ṭard al-ḥukm; see Astarābāḏī, Šārḥ, II, 63) as applied to words whose derivation is not known. Thus, the yāʾ of the appended word ǧayʾal, according to Ibn Ǧinnī (Munṣif, I, 35), can only be an augment in spite of the fact that the derivation of the word is unknown, because it can be demonstrated by examining other words that yāʾ or wāw can be one of the radicals (i.e., as opposed to augments) of quadriliterals only in reduplicated forms. Similarly, Astarābāḏī extends the rule through which the recurring consonant  is known to be an augment in a large number (kaṯīr) of words whose derivation is known—such as the appended words ḏuraḥriḥ, ḥiliblāb, and marmarīs, which he relates to their cognate triliteral roots—to those words whose derivation is not known—such as ṣamaḥmaḥ and barahraha—and clearly says that he does so by way of analogy so that the rule might be applicable to all attested examples (fa-ṭaradnā l-ḥukm fī l-kull; see Šārḥ, I, 63). As a result, the rule’s applicability is made to be universal rather than partial, and deviant examples become subject to the same rule that applies to the majority of the words of this type.
Turning to the identification of those words that do not qualify for inclusion in the closed system of appended material, it is clear that the grammarians not only strove to specify the characteristics of appended words so as to establish decisive criteria for the inclusion of material, but also dwelt on providing reasons for not including words that do not fit these criteria. As we saw earlier, particularly in 2.1–2.4, the grammarians specified the phonemes that may be used for ilḥāq and their positions within appended words, the number of radicals in these words as well as in the words to which they are appended, the major patterns into which they may be grouped, the nature of the relationship between the ziyāda of ilḥāq and that of meaning, the formal (lafẓī) rules that apply to the derivations of these words, and the suspension of the rule of idġām, where otherwise required, to them. Consequently, it may be said that any word that is at variance with any of these criteria cannot be part of the ilḥāq corpus. The following examples will demonstrate how non-appended words are identified by the application of these criteria and shed further light on the grammarians’ use of ilḥāq as a morphological testing device.
a. The position of the augment. Several rules are mentioned under this criterion (Astarābāḏī, Šarḥ, I, 56–57 and Suyūṭī, Hamʿ, II, 216–17). The hamza, for example, may be used for ilḥāq in medial and final positions, but in an initial position it cannot be an appending (mulḥiqa) augment unless it occurs together with another augment, referred to as musāʿid (aid). Thus, whereas ʾalandad and ʾidrawn are considered to be appended to safarǧal and ǧirdaḥl, respectively, because their initial hamza is accompanied by a nūn or a wāw, ʾafkal, ʾublum, and ʾiṯmid, whose initial hamza is the only augment, do not qualify as examples of ilḥāq .
b. The number of radicals. As was pointed out in 2.2 and 3.2 above, words were excluded from ilḥāq on the basis of the number of their radicals. The most obvious case here is that of augmented five-radical words, such as ʿaḍrafūṭ, ʿandalīb, qabaʿṯarā, and ḍabaġṭarā. Since there are no target words, i.e., six-radical words excluding any augment, to which these examples may be appended, the grammarians had to exclude them from the ilḥāq corpus and look elsewhere for possible interpretations. Hence, their explanation of the final alif in qabaʿṯarā—which they were also unable to explain as a feminine ending, since the word accepts nūnation and since the variant form qabaʿṯarāt does include the feminine ending (Sībawayhi, Kitāb, II, 9, 78, 342)—as an augment of enlargement (takṯīr al-kalima) may be viewed as one way out of a difficult problem that arose because of the limitations of ilḥāq (cf. Māzinī, Taṣrīf, I, 51, and Ibn ʿUṣfūr, Mumtiʿ, I, 206).
c. The “target” pattern. In addition to the lack of a target pattern to which augmented quinqueliterals may be appended (see “b” above), several other words, and even whole patterns, were not considered to be appended because of the lack of a target word or pattern to which they can be appended. For instance, Sībawayhi (Kitāb, I, 401–402) says that iḥmarartu and išhābabtu, both of triliteral roots, are not examples of ilḥāq because there is no quadriliteral of the type *iḥraǧamtu or *iḥrāǧamtu, respectively, to which they can be appended. Māzinī (Taṣrīf, II, 269) passes a similar judgment on iġdawdana for lack of the type *iḥrawǧama, as does Mubarrad (Muqtaḍab, IV, 3) with words like ʿaǧūz, raġīf, and risāla, which have no quadriliteral counterparts to which they can be appended. At times a whole pattern is said not to be intended for ilḥāq, as in the case of faʿlāʾ, for which there is no corresponding unhamzated pattern—i.e., a quadriliteral such as *sardāḥ or *sarbāl—to which it can be appended, and hence its two final alifs (i.e., ā and ʾ) are, according to Sībawayhi (Kitāb, II, 10) and Fārisī (Taʿlīqa, III, 38), used exclusively as a feminine ending.
d. The structure of the pattern. Contrary to “c” above, the target pattern may be available, but the structure of the words that can theoretically be appended to it prevent the process of ilḥāq. This may be illustrated by the pattern faʿlāl, which theoretically is a target pattern to which triliterals may be appended, but no triliteral was appended to it  because its examples are restricted to reduplicated biliterals (hence *faʿfāʿ), be they nouns (asmāʾ), such as zalzāl and ǧaṯǧāṯ, or adjectives (ṣifāt), such as ḥaṯḥāṯ and ḥaqḥāq (Sībawayhi, Kitāb, II, 338). The reason for this is that the structure of triliterals prevents the formation of reduplicated words, since this would theoretically require a nonexistent six-radical pattern. In comparison, the two sister patterns fiʿlāl and fuʿlāl were actually used as target patterns because their examples have four radicals that are not duplicated, such as qinṭār and qurṭās, and therefore words of triliteral origin like ǧilwāḫ and qurṭāṭ, respectively, lent themselves to be appended to them.
e. Meaning, derivatives and idġām. The discussion of these three criteria in 2.3 and 2.4 above included several examples of words and patterns that were considered, in each case, to be outside the sphere of ilḥāq because they do not conform to the criterion at hand.
4.1. As several examples cited above have shown, the various rules and details related to ilḥāq were used by the grammarians as a testing device for a host of morphological issues. So widespread was the practice that one may conclude that it represented for them a major objective, in addition to the principal objective of reducing the patterns within the closed system of appended words. Three of the most essential morphological premises they used ilḥāq as a testing device to check the validity of will be briefly discussed below.
a. The distinction between radicals according to aṣl and ziyāda. Appended words are used to confirm this distinction through the process of derivation and the realization of a common meaning they share with the roots. Sībawayhi (Kitāb, II, 116), for example, argues that ʿafarnā (strong lion), because of its affinity to ʿifr and ʿifrāt (both also mean “strong lion”), is an appended word because of the ziyāda of its n and ā, and he shows how this ziyāda—as well as that in ʿufāriya, which likewise means “strong lion”—is reflected in various aspects of their morphology. This is further tested by the four diminutive forms ʿufayrin, ʿufayrina, ʿufayr, and ʿufayriya, the first two of which prove that the ā of ʿafarnā is zāʾida, whereas the other two prove that its nūn is zāʾida. In this particular case, appended words are used to check the validity of the morphological rules that govern the diminutive and that are largely based on the distinction between what is aṣl and what is ziyāda in the words from which diminutives are formed.
b. The assignment of the position of the ziyāda. Since appended words mirror the phonological construction of the words that they are appended  to, including the positions of what is aṣl and what is ziyāda, they were used by the grammarians to check the correctness of the roots that they assign for augmented words. An example of this are the two words iḥranǧama and iḫranṭama, said by the grammarians to be quadriliterals (banāt al-arbaʿa) because they interpret the nūn as an augment (cf. Māzinī, Taṣrīf, I, 86). This interpretation is supported by the comparison some grammarians make (ibid., I, 86–89, and Ibn Ǧinnī’s commentary) between these words and appended words such as iqʿansasa and islanqā, the mawḍiʿ (position) of whose augmented nūn is determined to be between the ʿayn and the lām, i.e., the second and third original radicals of the roots QʿS and SLQ, respectively. It may thus be said that the two types of words reciprocally support the grammarians’ interpretation of each of them.
c. The identification of inadmissible patterns. The problem of identifying what is permissible and what is not seems to have occupied the earlier grammarians and lexicographers, probably as part of their effort to uncover the rules that determine the structure of Arabic words and consequently to be able to recognize as Arabicized or invented any word that is inconsistent with these rules. In this respect, the grammarians proposed several unattested patterns of ilḥāq which violate accepted structures in order to show that their impermissibility is due to the impermissibility of their counterparts to which they would have theoretically been appended. Ibn Ǧinnī’s (Munṣif, I, 88–89) masterly discussion of why patterns of the types *ifʿanwaltu, *ifʿanlaytu, *infanʿaltu, *īfanʿaltu do not occur reveals that these were proposed to demonstrate their incompatibility with the attested pattern ifʿanlaltu, as in iḥranǧamtu, to which augmented quadriliterals are usually appended (Suyūṭī, Muzhir, II, 41). Furthermore, the grammarians’ discussion of the criteria that disqualify words from being considered as examples of ilḥāq (see 3.3 above) shows how they repeatedly use them to check the validity of the rules that determine the permissibility or otherwise of target words and patterns to which other words and patterns may be appended.
The use of ilḥāq as a testing device nowhere finds it ultimate application  better than in what is known as masāʾil al-tamrīn (drill problems or exercises). Such drills are not only intended as pedagogical devices to train students and examine their grasp of the concepts involved. More essentially they test the applicability of these concepts through increasingly difficult questions, the answers to which should be in compliance with the theoretically permissible structures of Arabic words and patterns. Indeed, Ibn Ǧinnī, in his justification of Māzinī’s lengthy chapter dealing primarily with ilḥāq material and entitled “Analogically formed unsound words the only patterns of which are used in sound words” (hāḏā bāb mā qīsa min al-muʿtall wa-lam yaǧiʾ miṯāluhu illā min al-ṣaḥīḥ; Taṣrīf, II, 242–323), says that the reason for the invention of this “science” (ʿilm) is to use attested material as the basis for analogically constructing unattested material. By so doing, the grammarians could confirm the soundness of the morphological postulates that they used to explain attested usage. In Māzinī’s chapter, for example, the rules governing the use of wāw and yāʾ, known as iʿlāl, are thoroughly checked by arbitrary formulations such as *ibyayyaʿa, *uqwūwila, *iwʾawʾā, *īwīwāʾ, and ġazwawūt (II, 243, 245, 247, 251, 257, respectively). The fact that each of the questions which usually begin with the stereotype expression “Construe (ibni) x from y” should have one correct answer shows how the sum of rules that the grammarians deduced from usage worked together to yield attested words as well as theoretically usable words. Among the rules these drills seem to test in relation to wāw and yāʾ are the effect of vowels on them, the shift from one of them to the other, their compatibility and incompatibility, and principles related to gemination, omission, and their relation with hamza.
4.2. On a wider scale, the grammarians were keen to incorporate ilḥāq into their overall system of grammatical analysis and to demonstrate its pertinence to it beyond the morphological level. It is for this purpose that they try to show how some of their assumptions and general principles of analysis are harmonious with their approach to ilḥāq. An example of this is the principle that if a word is characterized by ṯiqal (heaviness), the  Arabs avoid the addition to it of another element that aggravates its ṯiqal, or introduce into it an element of ḫiffa (lightness) to counterbalance its ṯiqal. This very principle, which is especially familiar in naḥw—e.g., as in its application to justify the lack of tanwīn in diptotes (cf. Sībawayhi, Kitāb, I, 7), and the use ḍamma, due to its ṯiqal (heaviness), with the agent and the fatḥa, due to its ḫiffa (lightness), with the direct object, since a verb can have only one agent but may have more than one direct object (cf. Ibn al-Anbārī, Asrār, 78, and Baalbaki 1995, 87–88)—is carried over to the domain of ilḥāq. Thus, Ibn Ǧinnī (Munṣif, I, 51) argues that quinqueliterals were only augmented with one element (here, a long vowel) and not two elements since this would bring together two kinds of ṯiqal, that of the word’s structure and that of two augments. He then proceeds to show how this fact has direct bearing on ilḥāq because it limits the number of radicals that a target word can have. Astarābāḏī (Šārḥ, I, 64), on the other hand, invokes the principle of counterbalancing ṯiqal with ḫiffa to show that since it was not applied to the likes of mahdad and alandad—i.e., the ṯiqal of the augmented structures was not counterbalanced with the ḫiffa that gemination would have brought about—such words must have been intentionally deprived of gemination because they were meant to be appended words. Although this case is about the inapplicability of a particular principle to one kind of ilḥāq because of a compelling reason, the mere fact that it warranted such a justification is extremely important, since it demonstrates the expectation that ilḥāq not be at odds with other constituents of the grammatical system.
This expectation most probably owes its origin to the awareness of the grammarians that since they chose to treat ilḥāq as a distinct phenomenon to facilitate morphological analysis, although this was not dictated by the nature of the linguistic data (see 3.2 above), they had to defend its use and thus justify their choice. Obviously the most efficient way to do this was to demonstrate that ilḥāq is well accommodated to the general system and harmonious with some of it major principles, such as ḫiffa and ṯiqal, samāʿ, and qiyās (cf. Māzinī, Taṣrīf, I, 41; Ibn Ǧinnī, Ḫaṣāʾiṣ, I, 114), rejection of anomalous (šāḏḏ) data (cf. Astarābāḏī, Šārḥ, I, 69), resemblance to unattested material (Ibn Ǧinnī, Ḫaṣāʾiṣ, II, 343), and  the like, and that it operates according to well-defined rules that exhibit a logical relation among appended patterns (cf. Kitāb, II, 401, where Sībawayhi establishes the following correspondence: faʿall: faʿlal = fuʿull: fuʿlul = fiʿill: fiʿlil; and Ibn Ǧinnī, Munṣif, I, 47, where the relation between quinqueliterals and quadriliterals is said to be the same as that between quadriliterals and triliterals). Ultimately, perhaps, the grammarians wanted to demonstrate that ilḥāq is yet another proof of the underlying logic of language and to stress that it is the grammarian’s task is to discover the various ways in which this logic expresses itself.
A. Primary sources
Šarḥ = Raḍī l-Dīn Muḥammad b. Ḥasan
al-Astarābāḏī, Šarḥ Šāfiyat Ibn
al-Ḥāǧib. Ed. Muḥammad Nūr al-Ḥasan et al. 4 vols.
= Abū ʿAlī al-Ḥasan b. Aḥmad al-Fārisī, al-Masāʾil
al-muškila l-maʿrūfa bi-l-Baġdādiyyāt. Ed. ṣalāḥ
al-Dīn ʿAbd Allāh al-Sinkāwī.
———, Taʿlīqa = Abū
ʿAlī al-Ḥasan b. Aḥmad al-Fārisī, al-Taʿlīqa
ʿalā Kitāb Sībawayhi. Ed. ʿAwaḍ b. Ḥamad
al-Qūzī. 6 vols.
Farrāʾ, Maqṣūr = Abū Zakariyyāʾ Yaḥyā
b. Ziyād al-Farrāʾ, al-Maqṣūr wa-l-mamdūd. Ed. Māǧid al-ḏahabī,
Ǧawālīqī, Muʿarrab = Abū Manṣūr Mawhūb b. Aḥmad b. Muḥammad al-Ǧawālīqī, al-Muʿarrab min al-kalām al-aʿǧamī ʿalā ḥurūf al-muʿǧam. Ed. Aḥmad Muḥammad Šākir. Cairo: Dār al-Kutub al-Miṣriyya, 1361 A.H.
Ḫafāǧī, Šifāʾ = Šihāb al-Dīn Aḥmad b. Muḥammad al-Ḫafāǧī, Šifāʾ al-ġalīl fī-mā fī kalām al-ʿarab min al-daḫīl. Ed. Muḥammad Kaššāš. Beirut: Dār al-Kutub al-ʿIlmiyya, 1998.
Ḫalīl, ʿAyn = Abū ʿAbd al-Raḥmān al-Ḫalīl b. Aḥmad al-Farāhīdī, Kitāb al-ʿayn. Ed. Mahdī al-Maḫzūmī and Ibrāhīm al-Sāmarrāʾī. 8 vols. Baghdad: Dār al-Rašīd, 1980–85.
Ibn al-Anbārī, Asrār = Abū l-Barakāt ʿAbd al-Raḥmān b. Muḥammad al-Anbārī, Asrār al-ʿarabiyya. Ed. Muḥammad Bahǧat al-Bīṭār. Damascus: Maṭbaʿat al-Taraqqī, 1957.
Ibn ʿAqīl, Šarḥ = Bahāʾ al-Dīn ʿAbd Allāh Ibn ʿAqīl, Šarḥ Ibn ʿAqīl ʿalā Alfiyyat Ibn Mālik. Ed. Ramzī Munīr Baʿalbakī. Beirut: Dār al-ʿIlm lil-Malāyīn, 1992.
Ibn al-Dahhān, Šarḥ = Abū Muḥammad Saʿīd b. al-Mubārak b. ʿAlī Ibn al-Dahhān, Šarḥ Abniyat Sībawayhi. Ed. Ḥasan Šāḏilī Farhūd. Riyad: Dār al-ʿUlūm, 1987.
Ibn Durayd, Ǧamhara = Abū Bakr Muḥammad b. al-Ḥasan Ibn Durayd, Ǧamhrat al-luġa. Ed. Ramzī Munīr Baʿalbakī. 3 vols. Beirut: Dār al-ʿIlm lil-Malāyīn, 1987–88.
Ibn Ǧinnī, Ḫaṣāʾiṣ = Abū l-Fatḥ ʿUṯmān Ibn Ǧinnī, al-Ḫaṣāʾiṣ. Ed. Muḥammad ʿAlī al-Naǧǧār. 3 vols. Cairo: Dār al-Kutub al-Miṣriyya, 1952–56.
———, Munṣif = Abū l-Fatḥ ʿUṯmān Ibn Ǧinnī, al-Munṣif: Šarḥ Kitāb al-taṣrīf lil-Māzinī. Ed. Ibrāhīm Muṣṭafā and ʿAbd Allāh Amīn. 3 vols. Cairo: Muṣṭafā al-Bābī al-Ḥalabī, 1954–60.
———, Sirr = Abū l-Fatḥ ʿUṯmān Ibn Ǧinnī, Sirr ṣināʿat al-iʿrāb. Ed. Ḥasan Hindāwī. 2 vols. Damascus: Dār al-Qalam, 1985.
———, Taṣrīf = Abū l-Fatḥ ʿUṯmān Ibn Ǧinnī, al-Taṣrīf al-mulūkī. Ed. D‚esir‚e Saqqāl. Beirut: Dār al-Fikr al-ʿArabī, 1998.
Ibn al-Ḥāǧib, Šāfiya = Astarābāḏī, Šarḥ.
Ibn Hišām, Awḍaḥ = Ǧamāl al-Dīn Abū Muḥammad ʿAbd Allāh b. Yūsuf Ibn Hišām, Awḍaḥ al-masālik ilā Alfiyyat Ibn Mālik. Ed. Muḥammad Muḥyī l-Dīn ʿAbd al-Ḥamīd. 4 vols. Cairo: al-Maktaba al-Tiǧāriyya, 1956.
Ibn Kamāl Pasha, Risāla = Aḥmad b. Sulaymān Ibn Kamāl Pasha, Risāla fī taḥqīq taʿrīb al-kalima al-aʿǧamiyya. Ed. Muḥammad Sawāʿī. Damascus: Institut Fran«cais de Damas, 1991.
Ibn Mālik, Tashīl = Ǧamāl al-Dīn Abū ʿAbd Allāh Muḥammad b. ʿAbd Allāh Ibn Mālik, Tashīl al-fawāʾid wa-takmīl al-maqāṣid. Ed. Muḥammad Kāmil Barakāt. Cairo: Dār al-Kātib al-ʿArabī, 1967.
Ibn Manẓūr, Lisān = Ǧamāl al-Dīn Abū l-Faḍl Muḥammad b. Mukram Ibn Manẓūr, Lisān al-ʿarab. 20 vols. Būlāq, 1300–1307 A.H.
Ibn al-Sarrāǧ, Uṣūl = Abū Bakr Muḥammad b. Sahl Ibn al-Sarrāǧ, al-Uṣūl fī l-naḥw. Ed. ʿAbd al-Ḥusayn al-Fatalī. 3 vols. Beirut: Muʾassasat al-Risāla, 1985.
Ibn ʿUṣfūr, Mumtiʿ = Abū l-Ḥasan ʿAlī b. Muʾmin Ibn ʿUṣfūr al-Išbīlī, al-Mumtiʿ fī l-taṣrīf. Ed. Faḫr al-Dīn Qabāwa. 4th ed. 2 vols. Beirut: Dār al-Āfāq al-Ǧadīda, 1979.
Ibn Yaʿīš, Šarḥ al-Mufaṣṣal = Muwaffaq al-Dīn Yaʿīš b. ʿAlī Ibn Yaʿīš, Šarḥ al-Mufaṣṣal. 10 vols. Cairo: al-Maṭbaʿa al-Munīriyya, n.d.
———, Sarḥ al-Mulūkī = Muwaffaq al-Dīn Yaʿīš b. ʿAlī Ibn Yaʿīš, Šarḥ al-Mulūkī fī l-taṣrīf. Ed. Faḫr al-Dīn Qabāwa. Aleppo: al-Maktaba al-ʿArabiyya, 1973.
Māzinī, Taṣrīf = Ibn Ǧinnī, Munṣif.
Mubarrad, Muqtaḍab = Abū l-ʿAbbās Muḥammad b. Yazīd al-Mubarrad, al-Muqtaḍab. Ed. Muḥammad ʿAbd al-Ḫāliq ʿUḍayma. 4 vols. Cairo: Dār al-Taḥrīr, 1965–68.
Sībawayhi, Kitāb = Abū Bišr ʿAmr b. ʿUṯmān Sībawayhi, al-Kitāb. 2 vols. Būlāq, 1316–17 A.H.
Suyūṭī, Ašbāh = Ǧalāl al-Dīn Abū l-Faḍl ʿAbd al-Raḥmān b. Abī Bakr al-Suyūṭī, al-Ašbāh wa-l-naẓāʾir fī l-naḥw. 4 vols. Hyderabad: Maṭbaʿat Dāʾirat al-Maʿārif al-ʿUṯmāniyya, 1359–60 A.H.
———, Hamʿ = Ǧalāl al-Dīn Abū l-Faḍl ʿAbd al-Raḥmān b. Abī Bakr al-Suyūṭī, Hamʿ al-hawāmiʿ: Šarḥ Ǧamʿ al-ǧawāmiʿ fī ʿilm al-ʿarabiyya. 2 vols. Cairo: Maṭbaʿat al-Saʿāda, 1327 A.H.
———, Muzhir = Ǧalāl al-Dīn Abū l-Faḍl ʿAbd al-Raḥmān b. Abī Bakr al-Suyūṭī, al-Muzhir fī ʿulūm al-luġa wa-anwāʿihā. Ed. Muḥammad Aḥmad Ǧād al-Mawlā et al. 2 vols. Cairo: Dār Iḥyāʾ al-Kitāb al-ʿArabī, n.d.
Zamaḫšarī, Mufaṣṣal = Abū l-Qāsim Maḥmūd b. ʿUmar al-Zamaḫšarī, al-Mufaṣṣal fī ʿilm al-ʿarabiyya. Cairo, 1323 A.H.
Zubaydī, Amṯila = Abū Bakr Muḥammad b. al-Ḥasan al-Zubaydī, Amṯilat al-abniya fī Kitāb Sībawayhi. Ed. Muḥammad Ḫalīfa al-Danāʿ. Beirut: Dār al-Nahḍa al-ʿArabiyya, 1996.
Baalbaki, Ramzi. 1995. Teaching Arabic at University Level: Problems of Grammatical Tradition. In Proceedings of the Colloquium on Arabic Linguistics, Bucharest, August 29 – Sept. 2, 1994, ed. Nadia Anghelescu and Andrei A. Avram, pt. 1, 85–101. Bucharest: University of Bucharest.
———. 1998. Kitāb al-ʿAyn and Jamharat al-Lugha. In Early Medieval Arabic: Studies on al-Khalīl Ibn Aḥmad, ed. Karin C. Ryding, 44–62. Washington, D.C.: Georgetown University Press.
———. Theoretical Coherency versus Pedagogical Attainability: The Conscious Bias of Arab Grammarians. Forthcoming in a festschrift for Heinz Grotzfeld.
Bohas, Georges and Jean-Patrick Guillaume. 1984. Etude des th‚eories des grammairiens arabes. I. Morphologie et phonologie. Damascus: Institut Fran«cais de Damas.
Carter, Michael, G. 1973. An Arab Grammarian of the Eighth Century A.D. JAOS 93: 146–57.
———. 1981. The Use of Proper Names as a Testing Device in Sībawayhi’s Kitāb. HL 8, no. 2/3: 345–56.
Ḥadīṯī, Ḫadīǧa. 1965. Abniyat al-ṣarf fī Kitāb Sībawayhi. Baghdad: Maktabat al-Nahḍa.
Sara, Solomon I. 1991. Al-Khalīl: The First Arab Phonologist. IJIAS 8, no. 1: 1–57.
Talmon, Rafael. 1997. Arabic Grammar in its Formative Age: Kitāb al-ʿAyn and Its Attribution to Ḫalīl b. Aḥmad. Leiden : E. J. Brill.
ʿUḍayma, Muḥammad ʿAbd al-Ḫāliq. 1975. Fahāris Kitāb Sībawayhi. Cairo: Dār al-Ḥadīṯ.
 It should be noted that Ibn Ǧinnī, in his Sirr, generally avoids the use of the term ilḥāq, and uses the more general term ziyāda instead. For instance, he mentions a large number of the examples of ilḥāq of tāʾ, nūn (I, 167–69 for both) and wāw (II, 594) without referring to ilḥāq. He might have preferred ziyāda because it contrasts more directly with aṣl (i.e., what is part of the word’s root), since he tries to establish the contrast between what is augmented and what is a part of the root. Another possible reason is that since he investigates in his Sirr not only the morphological characteristics of the ḥurūf, but also their syntactic traits, and refers to the introduction of particles by using the root LḤQ (e.g., laḥāq, laḥiqat, talḥaq; Sirr, II, 325, 332, 384, 396, etc.), he consciously tried to avoid the term ilḥāq for the sense of appending, so as not to cause confusion between the two types.
 The meanings of the cited examples will be indicated only in cases where the semantic aspect is discussed, and the examples will be given mostly as nouns because the sources use them much more than verbs to illustrate ilḥāq.
 The phonemes of augment, including those used for ilḥāq, i.e., s,ʾ, l, t, m, w, n, y, h, ā, are generally referred to by mnemonic devices such as saʾaltumūnīhā, al-yawma tansāhu, hawītu l-simāna, wa-atāhu Sulaymān, amānun wa-tashīlun, taslīmun wa-hanāʾun, etc. See Ibn Ǧinnī, Munṣif, I, 98; Astarābāḏī Šārḥ, II, 331; Suyuṭī, Hamʿ, II, 214.
 See Bohas and Guilluame (1984, 109f.) for a discussion of ilḥāq by the addition of one of the letters of augment or by the recurrence of one of the roots, and the difference between the two types.
 We chose our two examples from loan words because they clearly have irreducible roots, but it must be noted that Arabic quadriliterals and quinqueliterals may also have irreducible roots, as in ǧaʿfar and ḫuzaʿbīla, both of which are mentioned as examples above. For other examples where the semantic aspect indicates the existence of ilḥāq, see Astarābāḏī, Šārḥ, II, 333 f. In certain cases, both the rules of augmentation and the semantic resemblance between the appended word and other derivatives from the same root point to the existence of ilḥāq. One example is kawṯar (abounding in good), whose wāw, according to Ibn Ǧinnī (Taṣrīf, 16), is an augment for two reasons, namely, that the word has three radicals other than the wāw, and that the meaning of abundance is present in the word kaṯīr, which is derived from the same root as kawṯar.
 The use of ilḥāq in the classification of words according to the number of their radicals should be distinguished from its use by some lexicographers for a similar classification into triliterals, quadriliterals and quinqueliterals. What a lexicographer like Ibn Durayd means by saying that certain words are “annexed” to the quinqueliterals (ulḥiqa bi-l-ḫumāsī) is that it is easier to classify them with the quinqueliterals as a distinct group, and not that they were made to conform to one of the patterns of the quinqueliterals as the more common use of the term ilḥāq implies. This explains why in the pattern fuʿālil, for example, dulāmiṣ and ǧuḫādib, considered by the grammarians to be triliterals appended to an augmented quadriliteral (see the second type mentioned in the text above), both appear in Ibn Durayd’s Ǧamhara (II, 1210, 1212) as examples of words that are “annexed” to quinqueliterals. In other words, Ibn Durayd is interested here in ilḥāq as a tool for classifying words in exhaustive lists, and not in the theoretical aspect of ilḥāq as discussed by the grammarians.
 Augmentation here mostly means the addition of a diphthong or a long vowel (ā, ū, or ī), probably since these, unlike short vowels, appear in writing.
 It is noteworthy that some grammarians use ilḥāq to distinguish between taṣrīf (morphology) and ištiqāq (derivation). Their argument is that the former is more general than the latter specifically because ilḥāq may be included under taṣrīf but not under ištiqāq (Suyūṭī, Muzhir, I, 351).
 This yāʾ refers to the written form of the word, and should be understood as a reference to the long vowel ī which was shortened to i (cf. zanādīq and zanādiqa).
 See Sībawayhi’s Kitāb, II, 197 and 211, for examples of the broken plurals of appended words.
 See the phonological reasons Bohas and Guillaume (1984, 39–41, 110–113) cite for this phenomenon.
 The same may be said of Mubarrad and Ibn al-Sarrāǧ (d. 316/929), who cite a large number of appended words (Muqtaḍab, see 1.2 above; and Uṣūl, esp. the chapter on abniya, III, 179–222) but do not cite any particular purpose for the phenomenon itself.
 This refers to words that can be either maqṣūr or mamdūd, such as fidā and fidāʾ, zinā and zināʾ, hayǧā, and hayǧāʾ (Farrāʾ, Maqṣūr, 38, 42, and 43, respectively).
 The final tāʾ of the word sanbata(tun), of course, should not count in the proposed pattern, otherwise its inclusion under faʿlal by Ibn al-Sarrāǧ would be inexplicable. He most probably included it under faʿlal because sanbat is its variant (Sībawayhi, Kitāb, II, 348; cf. 327; Ibn al-Dahhān, Šarḥ, 101; Astarābāḏī, Šarḥ, II, 340). Faʿlat is more likely to be intended by Ibn al-Sarrāǧ than faʿlan (see Suyūṭī, Muzhir, II, 15, for both possibilities) because the word after it, ʿansal, represents fanʿal, and Ibn al-Sarrāǧ systematically gives one example for each pattern.
 This grouping process which drastically reduces the number of “major” patterns is paralleled by the mostly Basran method of expressing augmented patterns, in most cases, by using only the letters f, ʿ, and l, in contrast with one Kufan method which allows the repetition of the same augments in the proposed pattern. Thus, safarǧal and šamardal are both represented as faʿallal according to the first method, but as faʿalǧal and faʿaldal, respectively, according to the second. Obviously, the first method avoids generating an exceedingly large number of patterns and readily reveals the words that belong to the same pattern. See Suyūṭī, Hamʿ, II, 213, for the differences among grammarians in expressing patterns; cf. Fārisī, Baġdādiyyāt, 529–31, and Astarābāḏī, Šārḥ, I, 10–21.
 For a study of this principle and its effect on the pedagogical attainability of grammatical rules, see Baalbaki (forthcoming). It should be mentioned here that since our primary sources on ilḥāq are almost exclusively Basran, we cannot say for certain whether there was a partisan divide on the issue or not, but the methods which the Basrans use in this case largely reflect their general interest in interpreting data in a way that would restrict the existence of deviations.
 In addition to this pattern, Sībawayhi refers to hammariš in two other places as being of the pattern faʿlalil (II, 341) and fanʿalil (II, 354). Cf. Lisān, HMRŠ, where Ibn Manẓūr attributes to Sībawayhi the proposal of two of these three patterns on two different occasions. See also Ibn ʿUṣfūr, Mumtiʿ, I, 269, and Suyūṭī, Muzhir, II, 29.
 As in the nouns wazwaza and waḥwaḥa (Taṣrīf, II, 216; III, 86), and yaʿyaʿa and yahyaha (Ibn Durayd, Ǧamhara, I, 216, 225).
 Astarābāḏī (II, 56), however, states that he finds no good reason why an initial hamza may not by itself, without a musāʿid, be considered mulḥiqa.
 Cf. n. 7 above. The word ḥandaqūq is usually mentioned with this group as well, but we did not include it because it is, as Ibn Ǧinnī rightly notes (Munṣif, I, 53), of a quadriliteral origin, since its qāf occurs twice.
 The earliest attempt of this kind is probably Ḥalīl’s introduction to Kitāb al-ʿAyn, written in the second half of the second century A.H. In it, he discusses some of the phonetic characteristics and phonotactics of Arabic words (I, 52–55) and specifically cites examples whose phonetic structure betray their foreign origin (e.g., duʿšūqa and ǧulāhiq) or their invention by skillful scholars (naḥārīr; e.g., kašaʿṯaǧ and ḫaḍaʿṯaǧ). See also Baalbaki 1998, 52–53, Sara 1991, 36–38, and Talmon 1997, 137–38.
 These drills are comparable to the grammarians’ practice of converting complex sentences into relative structures. The aim of this process, known as iḫbār (predication), as Carter (1981, 353) correctly argues, “may well have been to transform all utterances into propositions in order to test their truthfulness,” but it “finished up as a mere pedagogical device.” Likewise in the case of our drills, their pedagogical purpose has eventually gained supremacy over their use as a device for testing morphological rules.
 Surdad and sūdad, according to Ibn Ǧinnī, are appended to words that do not feature in actual usage but have the force of what is uttered (fī ḥukm al-malfūẓ). This is similar to the claim of the grammarians that some nouns, such as the interrogative particle mā, resemble supposed, non-existent particles that could be expected to have been coined but were not. Cf. Ibn ʿAqīl, Šārḥ, 32–33; Ibn Hišām, Awḍaḥ, I, 31; Baalbaki 1995, 87.