Triconsonantal Root Systems

The best topics from Languages & Linguistics, kept on a permanent basis.
Posts: 880
Joined: Fri Jul 03, 2009 3:17 pm
Location: Llundain

Re: Triconsonantal Root Systems

Post by Yng »


Perhaps a better way of looking at roots (which are not all tri-consonantal) is as something extracted from individual words during morphological processes and inserted into a new pattern to produce a new word, and as a convenient shorthand for the parts of the input word which remain constant through to the output word. It's better not to think of all words as a combination of a meaningful root and a meaningful pattern, because the former doesn't really exist, or at least not in any useful sense, and the latter requires us to postulate the existence of all sorts of patterns which may only occur with a small handful of words and have no clear derivational purpose. Not only this, but many words are derived from already-existing words and carry over an affixed or infixed consonant to the new word (ḥār > miḥwar > tamaḥwara for example), and some patterns convert vowels and semivowels into root consonants or only replicate part of the original word. Instead we should talk about derivational patterns as modifications of existing words and roots as the main thing that carries over in the course of that modification. Does that make sense? Derivation can happen from all parts of speech - there is denominal derivation, deverbal variation etc - but the important thing is that it doesn't happen directly from roots, because roots have no semantic value in and of themselves except inasmuch as they imply other words with that root. مكتب maktab 'office' and كاتب kātib 'clerk' are much more clearly directly related than maktab and some Platonic notion of 'writing'.


In Arabic there are somewhere between one and twenty verb derivation patterns of various levels of productivity. Ten of these are usually taught to learners as 'verb forms I-X' and correspond to the Binyanim of Hebrew, though only some of the Binyanim have cognates in Arabic. Although each form has common derivational semantics, because of semantic drift and other factors not every verb which can be classed as belonging to a form will show any clear relation to these semantics. Often the derivational link is muddied by the fact that like any other language, derivations can be somewhat idiomatic - قتل qatala 'kill' > قاتل qātala 'fight'. Furthermore, although each form has a typical ablaut pattern for the present and past stems and a typical form for the verbal noun, in this last category in particular there is considerable variation - تكلّم takallama's verbal noun, for example, should be تكلّم takallum, but in the vast majority of cases it is in fact كلام kalām. I'll sum up the usage and semantics of the ten forms people normally talk about here to give you an idea of how this morphology works and to show that it is as idiosyncratic and irregular as any other language.

Form I - which is the 'basic' or 'underived' form - has several different, largely randomly distributed patterns. There are some form I verbs that share the same root but differ in vowelling pattern and semantics, most typically on a basis that suggests that at least some of the variation in form I was originally a productive voice distinction - طلع ṭalaʿa 'climb up, go up' and طلع ṭaluʿa 'rise', for example. Form I verbs are characterised by having the shape CaCVC- as their preterite 'stem' (to which conjugation suffixes are added, I'll use the form 'stem' from here on).

Form II is used for several different purposes and is quite productive, especially in colloquial Arabic. It is formed by gemination of the middle root consonant, and one theory on its etymology is that it was originally a partial reduplication, or reduplication of monosyllabic/biconsonantal roots. It may originally have been predominantly used as an intensive of form I, and many form I-form II pairs are basically intensive or synonymous; the canonical example is قتل qatala 'kill' vs قتّل qattala 'slaughter'. It is also used, however, as a transitiviser and derives very broad causatives and transitives from verbs, especially in colloquial dialects where form IV has been basically lost: لبس labasa 'get dressed' > لبّس labbasa 'dress'. It also derives, from nouns, what some people call 'applicatives', verbs which give the meaning of applying some substance: ذهّب dhahhaba 'gild' < ذهب 'gold', and from adjectives verbs meaning 'find something X', 'consider something X' (نزّه على nazzaha ʿalā 'consider something too noble for' < نزيه nazīh 'noble'). These meanings are clearly linked semantically to the causative. I think it is cognate with Hebrew pi'el, which performs similar functions, although the connection has been somewhat muddied by the lose of gemination in modern Hebrew. Its verbnoun is formed, mysteriously, either on the pattern tafʿīl (e.g. taqtīl) or the pattern tifʿāl (e.g. tikrār 'repetition') - Hebrew has vestiges of what appears to be a causative or valency-increasing ti- in tif'el, which may be cognate to the verbnoun.

Form III is formed by the infixation of a long ā between the first and second consonants. Although we can identify derivational relationships between form III verbs and other verbs with the same root, form III is not very productive. Most of them express the process of 'striving for' something - قاتل qātala 'fight' does not in fact imply killing or even physical fighting, probably due to semantic drift, but is presumably derived from قتل qatala 'to kill' in this way originally. Other relationships are applicative in the typical cross-linguistic sense in that they change a prepositional argument into a core argument of the verb: كاتب kātaba 'write to' < كتب kataba ('write', which requires a preposition to express 'to').

Form IV is cognate to Hebrew hitpa'el and seems to cover pretty similar ground - causatives (ألغى alghā 'cancel' < لغا laghā 'be null') and verbs expressing a change of state (أشرق ashraqa 'rise in the East' < شرق sharq 'East', said of the sun, with a secondary and now more common meaning of 'shine' presumably derived from the former). Form IV has almost completely disappeared in colloquial dialects, retained only in a few learned classicisms, and has largely been collapsed into form I with its causative forms being taken over by their form II similes. It is relatively productive in MSA as a more high-register alternative to form II. Form IV has a cognate in Hebrew hif'il, and possibly previous to that shif'il (maybe sh- weakened to h-?).

Forms V and VI are most commonly the reflexive/reciprocal/passive forms of forms II and III and in this sense are very productive - in fact I'd say they border on being inflectional morphology rather than derivational. Their main formant is the addition of a prefix ta- to the aforementioned forms, and this ta- is probably cognate to the ta found in Hebrew hitpa'el as well as in Arabic's form VIII (these two are probably directly cognate). Some forms, thanks to semantic drift, have lost their specifically reciprocal element; تكلّم takallama 'speak' was probably originally the reciprocal of كلّم 'speak to', but they are now effectively synonymous. Forms V and VI - especially, productively, V - are also used to derive denominal verbs meaning 'act like' or 'pretend to be': تصامّ taṣāmma 'act deaf' < أصمّ aṣamm 'deaf', تأمّر taʾammara 'act imperiously' < أمير amīr 'prince'. They are also used - with a difference in aspect, probably originally deriving from the semantics of forms II and III - in the sense of 'become', with form III suggesting a slower, more gradual change: تصعّد taṣaʿʿada 'be escalated', تصاعد taṣāʿada 'slowly rise'.

Form VII's main formant is the prefix n-, and like V and VI, it is very productive in forming passives to the extent that it borders on being inflectional, especially in colloquial dialects. In North Africa, form VII exists as a passive but has the formant t-, possibly by analogy with forms V and VI or from some ancient dialectal variant. It forms almost exclusively straightforward passives of form I verbs: انقتل inqatala 'be killed' < قتل qatala 'kill'. Occasionally a form VII verb may act as a passive of a verb built on another measure.

Form VIII's main formant is a -t- infixed between the first and second consonants. Its primary function seems historically to have been a mediopassive of form I, and it is cognate with Hebrew hitpa'el, whose prefixed t- only becomes infixed with some kinds of verbs and in some forms (unlike Arabic's form VIII, where it is universally infixed). The t- is probably also cognate with that of forms V and VI and may have derived from a reflexive pronoun. It continues to form the mediopassive/reflexive of some verbs - احترق iḥtaraqa 'burn oneself' < حرق ḥaraqa 'burn'. In some cases it works as the mediopassive of a non-form I verb, such as استلم istalama 'receive' < سلّم sallama 'entrust, give to'. Form VIII also seems to be used to produce something that has been called the 'abstractive', which produces more 'metaphorical' versions of existing verbs: افتتح iftataḥa 'open [a meeting, a shopping centre]' < فتح fataḥa 'open [a door, a box]'. Many form VIIIs have no clear derivational connection with anything, presumably because of semantic drift.

Form IX is relatively rare and has disappeared in many colloquial dialects. It appears to be derived almost solely from adjectives formed on the pattern aCCaCu, which is limited to a relatively small semantic range usually summarised non-exhaustively as [the base] colours and [human] defects. Its distinctive formant is gemination of the final consonant, and it always gives the meaning 'be or become X', where X is the quality described by the adjective: احمرّ iḥmarra 'go red' < أحمر aḥmaru 'red'; اعمايّ iʿmāya 'go blind' < أعمى aʿmā 'blind' (the latter is very rare and slightly irregular). Almost all of the form IX verbs have at least one or two synonyms from the same root which are more commonly used.

Form X's main formant is the prefix sta-, and it is quite productive in a number of different functions, as well as having many verbs with no obvious derivational relationship to related verbs. One of its meanings is often summed up for non-linguists as 'seek to', which sums up a slightly complex shift in voice that converts a verb meaning 'do X' into a sort of suggestive causative, 'ask/try to get Y to do X'. One slightly obscure example of this is استقطع istaqṭaʿa 'ask Y to grant you X as a fief', from أقطع aqṭaʿa 'bestow as a fief'; a more obvious example is استعان istaʿāna 'seek help from' < أعان aʿāna 'help'. It also forms - particularly productively in some colloquial dialects - de-adjectival verbs meaning 'find something X, believe something to be X': استغرب istaghraba 'find something strange' < غريب. It also forms, from nouns or adjectives, verbs meaning 'become like X', usually with a volitional, human subject - so استغرب istaghraba also has a second meaning, this time probably derived from غرب gharb, which is 'to westernise onesself, to act western', whence مستغربون mustaghribūna 'Westernised people'. Another example is استرجل istarjala 'act like a man' < رجل rajul 'man'. A probably related and relatively niche meaning of form X participles, if not verbs - sometimes participle forms exist as nouns without verbal forms or with only hypothetical verbal forms - is the -ist of 'Arabist' and 'Orientalist' (مستعرب mustaʿrib and مستشرق mustashriq respectively, from عرب ʿarab 'Arabs' and شرق sharq 'East'). Hebrew has an apparent minor cognate in hishtaf'el, which appears to be derived from ship'el using the t- prefix; ship'el appears to be a precursor to the modern prefixed causatives in a- and ha- (Arabic and Hebrew respectively).

These are not the only verbal derivation patterns, but they received special attention from the Arab grammarians and then from Western grammarians because they most nicely exemplify the idea of a triliteral root. There are also many modifications of these patterns for deriving verbs from words with more than three consonants - تلفز talfaza 'televise' < تليفزيون televizyōn 'television', and many other patterns - particularly in colloquial dialects, but also in MSA, which are discarded in this analysis despite their clear etymological links to other words because they conjugate like normal four-letter verbs (but are actually derived from a three-letter root with the addition of another consonant in place of one of the root letters) - تولدن tawaldana 'act childish' < ولد walad 'child'. There are also large numbers of different, albeit related, patterns with some sort of sound symbolism - reduplication and so on - which produce iterative or intensive verbs, as well as onomatopeoia. There are many of these in MSA, such as معمع maʿmaʿa 'go baa [of sheep]' or ترجرج tarajraja 'sway, shake from side to side' from رجّ rajja 'shake', but there are many more in colloquial dialects and the patterns are relatively productive: لتلت latlat 'have a gossip < 'gossip' لتّ.


As far as substantives are concerned, there are a great number of nouns which do not seem to be formed on any particular pattern - including of course loanwords - and a few adjectives, too, but most adjectives fall into one of a few patterns, particularly CaCīC (e.g. صغير ṣaghīrun 'small', كبير kabīrun 'big'), etc. Many substantives are originally participles formed from verbs, largely (with the exception of the form I active participle) formed using the prefix mu- added to the verbal stem. Adjectives on the pattern CaCīC themselves, in fact, are very similar semantically to participles (equivalent to active participles for intransitive verbs of state/change of state, like صغر ṣaghira 'to become small', and passive participles for transitive verbs, like سجن sajana 'imprison', whence سجين sajīnun 'prisoner'); Aramaic uses this pattern productively for passive participles, and many Arabic verbs in practice use a CaCīC adjective in place of a participle. Another common category is what has often been called 'intensive' adjectives (a translation of the Arabic term), which are generally deverbal and imply habitual action: سكّير sikkīrun 'drunkard' < سكر sakira 'drink alcohol', مزواج mizwājun 'man given to marrying a lot' < تزوّج tazawwaja 'marry'. There are also nouns of place, formed with the prefix ma- or by substantivisation of the passive participle (مشفى mashfan 'hospital' < شفي shafiya 'cure'; or more commonly مستشفى mustashfan from استشفى with the same meaning) and nouns of instrument, generally formed with the prefix mi- (مفتاح miftāḥun 'key' < فتح fataḥa 'open') or on one of the intensive forms, usually in the feminine (فتّاحة fattāḥatun 'opener' < فتح fataḥa 'open'). These prefixes have become confused with one another and with the participle in colloquial dialects (and we can also see that the intensive and the noun of instrument seem to have always been conflated categories, perhaps unsurprisingly). Again, this is not an exhaustive list but it should be clear that ablaut morphology forms a very important part of the nominal derivation system. The only derivational morphology that does not involve any modification of the stem whatsoever, except for some borrowed suffixes from Turkish or European languages, is the 'relational' suffix -ī, which was originally used to form demonyms only but now is used as a generic adjective formant: صحافي ṣaḥāfiyyun 'journalist' < صحافة ṣaḥāfatun 'press'. But even this has a variant which imposes a pattern: صحفي ṣaḥafiyyun, or مدني madaniyyun 'civilian' < مدينة madīnatun 'city'.

On another kind of substantive: every verb in theory has a verbal noun, which generally, although not universally, is predictable for dimorphemic+ verbs (i.e. verbs not in form I) but is not predictable for form I. There is a wide range of different verbal noun patterns. Two - CaCCatun and CiCCatun, the 'noun of instance' and 'noun of manner' (the latter of which is no longer productive) have very obvious meanings: the former indicates one instance of the action described by the verb (ضربة ḍarbatun 'a [single] blow') and the latter translates 'way of' (ميتة miitatun '[noble, ignoble, sad etc] death'). Others are far less predictable; there are several common patterns (CaCaC, CaCC, CuCūC etc) which have no obvious semantics. Often verbs have two or more possible verbal nouns, one of which may (generally arbitrarily) have different semantics to the other. A few patterns have semantics you can at least vaguely pin down - CaCC3ūC3atun for example - seem to express a durative process/state: صيرورة ṣayrūratun 'development, process of becoming', قيلولة qaylūlatun 'siesta'. CuCāCun - often described as being a 'noun of illnesses' - seems to be rather a kind of resultative describing the state of the object of the verb post-action: زكام zukāmun 'cold', ركام rukāmun 'pile', etc. Some, like CaCāCatun, are used to derive abstract nouns from stative verbs/adjectives (سعادة saʿādatun 'happiness').

As for inflectional morphology, you're right that a significant part of it is achieved by means of suffixes and prefixes. As mentioned above, most participles are formed by the addition of a prefix mu- to the verb stem. Most adjectives and nouns form their feminine singular through addition of a suffix -atun, and their feminine plural through addition of a suffix -ātun (some foreign animate nouns also form their masculine plural this way: باشا bāshā 'Pasha' > باشاوات bāshawāt 'Pashas', as do verbal nouns and many other kinds of nouns). Many nouns and most adjectives also form their masculine plurals through the addition of a suffix -ūna, and all nouns and adjectives - without exception - form their dual through the addition of a suffix -āni. All singular forms and broken plurals express case morphology through the addition of suffixes and definiteness through a combination of prefixes and suffixes (indefinite nouns and adjectives generally take a -n suffix after the case suffix, definite adjectives take the l- prefix). Verbs, on the other hand - although there are usually differences in the internal vowelling of the stem between present and past - take suffixes to mark person, number and gender in the past and a combination of prefixes and suffixes to mark person, number, gender and mood in the present.

However, most nouns and a number of adjectives have what are called 'broken plurals', which is to say that they form their plural through root-and-pattern morphology; this is not the case in Hebrew, where all nouns and adjectives form the plural through suffixial morphology. There are a great number of plural patterns, many of which are productively applied to new borrowings and coinages (فيلم film > أفلام aflāmun 'films'); they are largely assigned to nouns based on the shape of the singular (aCCāCun is applied almost without exception to words with the shape CVCC or CVCVC). Some adjectives of specific patterns - the aforementioned aCCaCu colour/defect nouns, comparatives as well as adjectives formed with the pattern CaCCānu (but not its variant CaCCānun, which is spelt identically but has a different declension pattern and forms its feminine and plural with suffixes), also have a feminine formed by modification of the stem: أحمر aḥmaru 'red [m]' > حمراء ḥamrāʾu 'red [f], أكبر akbaru 'bigger [m]' > كبرا kubrā 'bigger [f]', تعبان taʿbānu 'tired [m] > تعبى taʿbā 'tired [f]'). Adjectives' comparatives are derived using root-and-pattern morphology on the pattern aCaCCu. Verbs, too - although most of their conjugation takes place through affixial morphology - have an internal passive which is formed through a change in vowels, presumably cognate to Hebrew's passive binyanim, which can be applied in theory to any verb. So root-and-pattern morphology plays an important role in inflection too. Inasmuch as participles and verbal nouns are inflectional rather than derivational morphology, they often also involve root-and-pattern modifications.

on ur other questions


'state' is a semiticist term which is possibly translated from Arabic grammar, who knows (I don't think so though), and basically refers to a morphological category of nouns and adjectives which is generally described as being 'absolutive', 'emphatic', 'construct'. Arabic has three of these - absolutive, definite, construct. 'Construct' is the only one I see commonly used in modern Arabic grammar, with 'absolutive' and 'emphatic' replaced by indefinite and definite.

The absolutive or indefinite state is marked, on most nouns, by the presence of 'nunation', a -n suffix; this is part of Arabic's case morphology and is not found in Aramaic or Hebrew except in the plural endings (in Hebrew the cognate is -m). Note that in all of these languages the masculine sound plural suffix, īm in Hebrew, īn in Aramaic and ūna/īna (varying for case) in Arabic, always has the nasal whether the noun is absolutive or definite, as does the dual; it is only in singular nouns that it is a marker of the absolutive (although this may not always have been the case). Absolutive nouns are generally semantically indefinite because they lack the definite article, although many proper names that lack the article may be absolutive but definite - e.g. Muḥammadun. There are also absolutive nouns that, because of their declension class, take no -n, like مكاتب makātibu 'libraries'.

The emphatic or definite state, in Arabic, is marked by the presence of a definite prefix al- or ha- in Hebrew. Adjectives agree for definiteness with their nouns and the prefix assimilates to the initial letter in Arabic, which are both good reasons for calling it a prefix rather than an article per se.

The construct state is essentially when the noun is neither absolutive nor indefinite. In the vast majority of cases, this is because it has been placed in a relationship of possession (or something similar, the actual semantic relationships covered by this structure are numerous) with a following noun or cliticised possessive pronoun. In Arabic the changes to a noun in construct state are relatively straightforward; the indefinite final -n is lost, as are the -ns of the plural and dual endings, and in five unusual nouns the case vowel is lengthened (أب ab 'father' > أبو abū 'father of'). Hebrew, as I understand it, likewise loses the final -n but also has various different shortenings and reductions which are applied to the possessed word (insertion/shortenings/loss of vowels and stress shifts in the construct state are also characteristic of many colloquial Arabic dialects). The possessor is placed in the genitive and will be definite or indefinite depending on its role in the sentence.


Classical Arabic, supposedly, does not really have a category of adverb; rather it creates adverbial structures using accusatives. Arabic is full of these. Most of its prepositional structures are actually locative nouns: عقب ʿaqiba 'just after' < 'heel', for example. Likewise, أتى متأخرًا atā mutaʾakkhiran 'he came late', where mutaʾakkhiran is the accusative indefinite of 'late'. In modern Arabic, adverbs are derived freely with the accusative indefinite suffix -an, even in colloquial dialects where case has been lost (-an has been reinterpreted as an adverbial). Other than this there are no productive adverb derivations, adjectives serve as adverbs unmarked much of the time in colloquial Arabic.

Articles, conjunctions, pronouns

articles and pronouns are not declined - there's no independent article, only a prefix, which is invariable (in both Hebrew and Arabic). In Arabic at least the independent pronoun, which is invariable, rarely appears except to introduce a topic or to emphasise or disambiguate an already-encoded pronoun, which will either appear as an attached clitic (which stands in for accusative and genitive nouns) or a flexional affix (which stands in for a nominative noun). Of the other kinds of pronouns - demonstrative, relative etc - only the dual relative pronoun declines to agree with its noun, probably because it has the normal dual suffix and of necessity usually appears after another dual suffix with exactly the same form and is otherwise similar to an adjective (it takes definite agreement, etc). Conjunctions and prepositions are not really prefixes - some of them are monosyllables and written attached to the word, but syntactically they are clitics and some of them can be pronounced independently (فـ fa- 'so' comes to mind) and basically act as independent words. Hebrew has a relativising conjunction which attaches itself to the first word of the relative clause, and some Arabic dialects - Palestinian and Lebanese off the top of my head - can also attach their relative pronoun illi to the first noun of a relative clause, as il-. But conjunctions are generally like English.
Last edited by Yng on Fri Oct 02, 2015 8:35 am, edited 3 times in total.
كان يا ما كان / يا صمت العشية / قمري هاجر في الصبح بعيدا / في العيون العسلية

tà yi póbo tsùtsùr ciivà dè!

short texts in Cuhbi

Risha Cuhbi grammar

Posts: 2
Joined: Mon Jul 28, 2014 8:53 pm

Re: Triconsonantal Root Systems

Post by DLJ393 »

Thank you for the reply. I appreciate very much the thoroughness of your explanations; my understanding of triconsonantal root systems was (and still is) very superficial.

I am not trained as a linguist or grammarian; it will take me a couple of readings to absorb all that you have presented.

Thanks again for this material; I am enjoying learning something new!

Post Reply