Shilha language

Tifinagh (or rather, Neo-Tifinagh) was introduced in the late 1990s and its use is now supported by the Moroccan authorities, in a standardised form promulgated by the Institut Royal de la Culture Amazighe (IRCAM, Rabat). Publications entirely printed in Tifinagh still remain rare, and only a tiny proportion of Shilha speakers, if any, are able to handle the new script with confidence. Its main role is emblematic, that of a cultural icon. As such, Tifinagh has entered the public space, with town signs now showing the name in Tifinagh as well as in Arabic and Latin script

Many Shilha texts from the oral tradition have been published since the 19th century, transcribed in Latin script. Early publications display a wide variety of transcription systems. Stumme (1899) and Destaing (1920, 1940) use an elaborate phonetic transcription, while Justinard (1914) and Laoust (1936) employ a transcription based on French orthographical conventions. A new standard was set by Aspinion (1953), who uses a simple but accurate, largely phonemic transcription with hyphenation.

Most academic publications on Berber languages of recent decades use a version of the Berber Latin alphabet, a transcription orthography in Latin script (as used in this article).[12] The most unusual feature of this orthography is the employment of the symbol ⟨ɛ⟩ (Greek epsilon) to represent /ʢ/ (voiced epiglottal fricative); for example, taɛmamt /taʢmamt/ "turban". Except with ⟨ḥ⟩ (= IPA /ʜ/), the subscript dot indicates pharyngealisation; for example, aḍrḍur /adˤrdˤur/ "deaf person". Geminated and long consonants are transcribed with doubled symbols, for example, tassmi "needle", agg°rn "flour". Word divisions are generally disjunctive, with clitics written as separate words (not hyphenated).

Shilha written in Arabic script: an 18th-century manuscript of al-Ḥawḍ by Mḥmmd Awzal.

Traditional orthography

Traditional Shilha manuscript texts are written in a conventionalized orthography in Maghribi Arabic script.[13] This orthography has remained virtually unchanged since at least the end of the 16th century, and is still used today in circles of traditional Islamic scholars (ṭṭlba). The main features of the traditional orthography are the use of two extra letters (kāf with three dots for g, and ṣād with three dots for ẓ) and full vocalization (vowels written with fatḥah, kasrah and ḍammah). Clitical elements are written connected to a noun or verb form (conjunctive spelling).[14]

Modern orthography

Since the 1970s, a fair number of books in Shilha have been published inside Morocco, written in a newly devised, practical orthography in Arabic script. The main features of this orthography are the representation of vowels a, i, u by the letters alif, yāʼ, wāw, and the non-use of vocalization signs other than shaddah (to indicate gemination of consonants) and ḍammah (to indicate labialization of velar and uvular consonants). The consonant g is written with گ, and ẓ is either written with زٜ (zāy with dot below) or not distinguished from z. Word separations are mostly disjunctive.

There is currently no evidence of word stress in Tashlhiyt.[16]

Shilha has three phonemic vowels, with length not a distinctive feature.[17] The vowels show a fairly wide range of allophones.[18] The vowel /a/ is most often realized as [a] or [æ], and /u/ is pronounced without any noticeable rounding except when adjacent to w. The presence of a pharyngealized consonant invites a more centralized realization of the vowel, as in kraḍ [krɐdˤ] "three", kkuẓ [kkɤzˤ] "four", sḍis [sdˤɪs] "six" (compare yan [jæn] "one", sin [sin] "two", smmus [smmʊs] "five").

Additional phonemic vowels occur sporadically in recent loanwords, for example /o/ as in rristora "restaurant" (from French).

In addition to the three phonemic vowels, there are non-phonemic transitional vowels, often collectively referred to as "schwa". Typically, a transitional vowel is audible following the onset of a vowelless syllable CC or CCC, if either of the flanking consonants, or both, are voiced,[19] for example tigmmi [tiɡĭmmi] "house", amḥḍar [amɐ̆ʜdˤɐr] "schoolboy". In the phonetic transcriptions of Stumme (1899) and Destaing (1920, 1940), many such transitional vowels are indicated.

Later authors such as Aspinion (1953), use the symbol ⟨e⟩ to mark the place where a transitional vowel may be heard, irrespective of its quality, and they also write ⟨e⟩ where in reality no vowel, however short, is heard, for example ⟨akessab⟩ /akssab/ "owner of livestock", ⟨ar icetta⟩ /ar iʃtta/ "he's eating". The symbol ⟨e⟩, often referred to as "schwa", as used by Aspinion and others, thus becomes a purely graphical device employed to indicate that the preceding consonant is a syllable onset: [a.k(e)s.sab], [a.ri.c(e)t.ta].[20] As Galand has observed, the notation of "schwa" in fact results from "habits which are alien to Shilha".[21] And, as conclusively shown by Ridouane (2008), transitional vowels or "intrusive vocoids" cannot even be accorded the status of epenthetic vowels. It is therefore preferable not to write transitional vowels or "schwa", and to transcribe the vowels in a strictly phonemic manner, as in Galand (1988) and all recent text editions.[lower-alpha 3]

The chart below represents Tashlhiyt consonants in IPA, with orthographical representations added between angled brackets when different:[22]

Additional phonemic consonants occur sporadically in recent loanwords, for example /bʷ/ as in bb°a "(my) father" (from Moroccan Arabic), and /p/ as in laplaj "beach" (from French).

Like other Berber languages and Arabic, Tashlhiyt has both pharyngealized ("emphatic") and plain dental consonants.[23][24] There is also a distinction between labialized and plain dorsal obstruents.[24] Consonant gemination or length is contrastive.[25]

The semivowels /w/ and /j/ have vocalic allophones [u] and [i] between consonants (C_C) and between consonant and pause (C_# and #_C). Similarly, the high vowels /u/ and /i/ can have consonantal allophones [w] and [j] in order to avoid a hiatus. In most dialects,[26] the semivowels are thus in complementary distribution with the high vowels, with the semivowels occurring as onset or coda, and the high vowels as nucleus in a syllable. This surface distribution of the semivowels and the high vowels has tended to obscure their status as four distinct phonemes, with some linguists denying phonemic status to /w/ and /j/.[27]

Positing four distinct phonemes is necessitated by the fact that semivowels and high vowels can occur in sequence, in lexically determined order, for example tazdwit "bee", tahruyt "ewe" (not *tazduyt, *tahrwit). In addition, semivowels /w/ and /j/, like other consonants, occur long, as in afawwu "wrap", tayyu "camel's hump".[28] The assumption of four phonemes also results in a more efficient description of morphology.[29]

In the examples below, w and y are transcribed phonemically in some citation forms, but always phonetically in context, for example ysti- "the sisters of", dars snat istis "he has two sisters".

Any consonant in Tashlhiyt, in any position within a word, may be simple or geminate. There may be up to two geminates in a stem, and up to three in a word.[30]

The role of gemination varies:[30]

Gemination also may occur due to phonological assimilation. For example, the following phrase would be realized as [babllfirma]:[30]

Some consonants are realized differently during morphological gemination; ḍḍ as ṭṭ, ww as gg°, ɣɣ as qq, and ɣɣ° as qq°.[30]

Shilha syllable structure has been the subject of a detailed and highly technical discussion by phoneticians. The issue was whether Shilha does or does not have vowelless syllables. According to John Coleman, syllables which are vowelless on the phonemic level have "schwa" serving as vocalic nucleus on the phonetic level. According to Rachid Ridouane on the other hand, Shilha's apparently vowelless syllables are truly vowelless, with all phonemes, vowels as well as consonants, capable of serving as nucleus. The discussion is summed up in Ridouane (2008, with listing of relevant publications), where he conclusively demonstrates that a perfectly ordinary Shilha phrase such as tkkst stt "you took it away" indeed consists of three vowelless syllables [tk.ks.tst:.], each made up of voiceless consonants only, and with voiceless consonants (not "schwa") serving as nucleus. Many definitions of the syllable that have been put forward do not cover the syllables of Shilha.[31]

The syllable structure of Shilha was first investigated by Dell and Elmedlaoui in a seminal article (1985). They describe how syllable boundaries can be established through what they call "core syllabification". This works by associating a nucleus with an onset, to form a core syllable CV or CC. Segments that are higher on the sonority scale have precedence over those lower on the scale in forming the nucleus in a core syllable, with vowels and semivowels highest on the scale, followed by liquids and nasals, voiced fricatives, voiceless fricatives, voiced stops and voiceless stops. When no more segments are available as onsets, the remaining single consonants are assigned as coda to the preceding core syllable, but if a remaining consonant is identical to the consonant that is the onset of the following syllable, it merges with it to become a long consonant. A morpheme boundary does not necessarily constitute a syllable boundary.

Comparative diagram of the following:

Application of core syllabification produces the following Shilha syllable types:

Shilha syllable structure can be represented succinctly by the formula CX(C), in which C is any consonant (single/long), and X is any vowel or consonant (single) and with the restriction that in a syllable CXC the X, if it is a consonant, cannot be higher on the resonance scale than the syllable-final consonant, that is, syllables such as [tsk.] and [wrz.] are possible, but not *[tks.] and *[wzr.].

Exceptional syllables of the types X (vowel or single/long consonant) and V(C) (vowel plus single/long consonant) occur in utterance-initial position:

rgl t [r.glt.] "close it!" (syllable C)

ffɣat [f:.ɣat.] "go out!" (syllable C:)

awi t id [a.wi.tid.] "bring it here!" (syllable V)

acki d [ac.kid.] "come here!" (syllable VC)

Another exceptional syllable type, described by Dell and Elmedlaoui (1985), occurs in utterance-final position, when a syllable of the type CC or CC: is "annexed" to a preceding syllable of the type CV or C:V, for example fssamt "be silent!" is [fs.samt.] not *[fs.sa.mt.].

Since any syllable type may precede or follow any other type, and since any consonant can occur in syllable-initial or final position, there are no phonotactical restrictions on consonant sequences. This also means that the concept of the consonant cluster is not applicable in Shilha phonology, as any number of consonants may occur in sequence:

[fr.ḥɣs.lm.ɛrf.tn.nk.] (6 syllables, 14 consonants, no vowels)

The metrics of traditional Shilha poems, as composed and recited by itinerant bards (inḍḍamn), was first described and analyzed by Hassan Jouad (thesis 1983, book 1995; see also Dell and Elmedlaoui 2008). The traditional metrical system confirms the existence of vowelless syllables in Shilha, and Jouad's data have been used by Dell and Elmedlaoui, and by Ridouane to support their conclusions.

The metrical system imposes the following restrictions:

• each line in a poem contains the same number of syllables as all the other lines
• each syllable in a line contains the same number of segments as its counterpart in other lines
• each line contains one particular syllable that must begin or end with a voiced consonant
• each line is divided into feet, with the last syllable in each foot stressed ("lifted") in recitation

Within these restrictions, the poet is free to devise his own metrical form. This can be recorded in a meaningless formula called talalayt which shows the number and the length of the syllables, as well as the place of the obligatory voiced consonant (Jouad lists hundreds of such formulae).

The system is illustrated here with a quatrain ascribed to the semi-legendary Shilha poet Sidi Ḥammu (fl. 18th century) and published by Amarir (1987:64):

a titbirin a tumlilin a timgraḍ
ab bahra wr takkamt i lxla hann lbaz
igan bu tassrwalt ig lxatm ɣ uḍaḍ
ak k°nt yut ukan iɣli d ignwan izug

"O white doves, O pets!
Do not venture into the desert too often, for there is the falcon,
Wearing small trousers; he'll put a ring on [your] finger,
To strike you but once — then he ascends into the sky and is gone!"

Application of Dell and Elmedlaoui's core syllabification reveals a regular mosaic of syllables:

The poem is composed in a metre listed by Jouad (1995:283) and exemplified by the formula a láy, la li la láy, la li la lá, li lád (the d in the last syllable indicates the position of the compulsory voiced consonant).

On the basis of their morphology, three types of Shilha nouns can be distinguished, two indigenous types and one type of external origin:

• inflected nouns
• uninflected nouns
• unincorporated loans

The relevant morpho-syntactic categories are gender, number and state.[32]

Inflected nouns are by far the most numerous type. These nouns can be easily recognised from their outward shape: they begin with a nominal prefix which has the form (t)V-:

azal "daytime"

igigil "orphan"

uṣkay "hound"

tadgg°at "evening"

tibinṣrt "marsh mallow (plant)"

tuḍfit "ant"

Inflected nouns distinguish two genders, masculine and feminine; two numbers, singular and plural; and two states, conventionally referred to by their French names as état libre ("free state") and état d'annexion ("annexed state")[33] and glossed as EL and EA. Gender and number are all explicitly marked, but historical and synchronic sound changes have in some cases resulted in the neutralization of the difference between EL and EA.

The nominal prefix has no semantic content, i.e. it is not a sort of (in)definite article, although it is probably demonstrative in origin. It is made up of one or both of two elements, a gender prefix and a vocalic prefix. Singular feminine nouns may also have a gender suffix. For example, the noun tazdwit "bee" has the feminine prefix t-, the vocalic prefix a- and the feminine singular suffix -t added to the nominal stem zdwi. While feminine inflected nouns always have the feminine prefix, masculine nouns do not have a gender prefix in the free state (EL); for example abaɣuɣ "fox" has no gender prefix, but only a vocalic prefix a- added to the nominal stem baɣuɣ.

Gender is thus marked unambiguously, albeit asymmetrically. In just a handful of nouns, the morphological gender does not conform to the grammatical gender (and number): ulli "sheep and goats" is morphologically masculine singular, but takes feminine plural agreement; alln "eyes" is morphologically masculine plural, but takes feminine plural agreement; tarwa "(someone's) children, offspring" is morphologically feminine singular, but takes masculine plural agreement.

The annexed state (EA) is regularly formed by reducing the vocalic prefix to zero and, with masculine nouns, adding the masculine gender prefix w-:[34]

EL t-a-zdwi-t "bee" → EA t-zdwi-t

EL a-baɣuɣ "fox" → EA w-baɣuɣ

With some nouns, the original vocalic prefix has fused with a stem-initial vowel, to produce an inseparable (and irreducible) vowel:

EL ayyur "moon, month" → EA w-ayyur (not *w-yyur)

EL t-afuk-t "sun" → EA t-afuk-t (not *t-fuk-t)

With feminine nouns that have an inseparable vocalic prefix, the difference between EL and EA is thus neutralized.

While most inflected nouns have a vocalic prefix a-, some have i- (in some cases inseparable), and a few have u- (always inseparable). When a masculine noun has the vocalic prefix i- (separable or inseparable), the masculine gender prefix w- changes to y-. The table below presents an overview (all examples are singular; plurals also distinguish EL and EA):

The EA is not predictable from the shape of the noun, compare:

afus "hand" → EA wfus

afud "knee" → EA wafud

The phonological rules on the realization of /w/ and /j/ apply to the EA as well. For example, the EA of a-mɣar "chief" is /w-mɣar/, realized as wmɣar after a vowel, umɣar after a consonant:

idda wmɣar s dar lqqaḍi "the chief went to see the judge"

imun umɣar d lqqaḍi "the chief accompanied the judge"

Inflected nouns show a great variety of plural formations, applying one or more of the following processes:

• suffixation (masculine -n, feminine -in)
• vowel change (insertion or elision, or ablaut)
• consonant gemination or degemination
• stem extension (+aw, +iw, +t, +w, always in combination with a suffix)

There are also irregular and suppletive plurals. The feminine singular suffix -t is naturally lost in the plural.

Independent from these processes, the separable vocalic prefix a- is always replaced with i-. An inseparable vocalic prefix either remains unchanged, or changes as part of vowel change (but if the vocalic prefix is inseparable in the singular, it may be separable in the plural, as with aduz "dune", and vice versa, as with aydi "dog"; see table below).

Below is a sample of nouns, illustrating various plural formations.

The plural is generally not predictable from the shape of the singular, compare:

aduku "shoe", plural idukan (vowel change and suffix)

aruku "utensil", plural irukutn (stem extension and suffix)

Many nouns have more than one plural, for example a-jnwiy "knife", plural i-jnway (vowel change) or i-jnwiy-n (suffixation).

Many Shilha place-names are morphologically inflected nouns:

A-nammr "Anammeur"

I-ɣ°rays-n " Irhoreïsene"

T-a-rudan-t "Taroudant"

T-i-zgzaw-in "Tizegzaouine"

The same is the case with Shilha ethnic names:

Amml-n "the Ammeln" (singular Imml)

Actuk-n "the Achtouken" (singular Actuk)

I-lall-n "the Ilallen" (singular I-lillu)

I-sk°ta-n "the Isouktan" (singular A-sktu)

Among the inflected nouns are found many incorporated loans. Examples include:

t-a-kira "wax" (from Latin)

a-ɣanim "reeds" (from Punic)

urti "vegetable plot, orchard" (from early Romance)

a-muslm "Muslim" (from Arabic)

t-a-bra-t "letter, missive" (from Arabic)

This is the least common type, which also includes some loans. Examples:

dikkuk "cuckoo"

fad "thirst"

gmz "thumb"

kḍran "tar" (from Arabic)

lagar "station" (from French)

mllɣ "index finger"

sksu "couscous"

wiẓugn "cricket"

xizzu "carrots"

It is probable that all uninflected nouns were originally masculine. The few that now take feminine agreement contain elements that have been reanalyzed as marking feminine gender, for example ttjdmnni "kind of spider" (initial t seen as feminine prefix), hlima "bat" (not an Arabic loanword, but final a analyzed as the Arabic feminine ending).

Many uninflected nouns are collectives or non-count nouns which do not have a separate plural form. Those that have a plural make it by preposing the pluralizer id, for example id lagar "stations".

The uninflected noun mddn or middn "people, humans" is morphologically masculine singular but takes masculine plural agreement.

Names of people and foreign place-names can be seen as a subtype of uninflected nouns, for example Musa (man's name), Muna (woman's name), Fas "Fès", Brdqqiz "Portugal". Gender is not transparently marked on these names, but those referring to humans take gender agreement according to the natural sex of the referent (male/masculine, female/feminine).

These are nouns of Arabic origin (including loans from French and Spanish through Arabic) which have largely retained their Arabic morphology. They distinguish two genders (not always unambiguously marked) and two numbers (explicitly marked). A notable feature of these nouns is that they are borrowed with the Arabic definite article, which is semantically neutralized in Shilha:

Moroccan Arabic l-kabus "the pistol" → Shilha lkabus "the pistol, a pistol"

Moroccan Arabic t-tabut "the coffin" → Shilha ttabut "the coffin, a coffin"

The Arabic feminine ending -a is often replaced with the Shilha feminine singular suffix -t:

Moroccan Arabic l-faky-a → Shilha lfaki-t "fruit"

Moroccan Arabic ṛ-ṛuḍ-a → Shilha ṛṛuṭ-ṭ "tomb of a saint"

Arabic loans usually retain their gender in Shilha. The exception are Arabic masculine nouns which end in t; these change their gender to feminine in Shilha, with the final t reanalyzed as the Shilha feminine singular suffix -t:

Moroccan Arabic l-ḥadit "the prophetic tradition" (masculine) → Shilha lḥadi-t (feminine)

Moroccan Arabic l-mut "death" (masculine) → Shilha lmu-t (feminine)

Arabic plurals are usually borrowed with the singulars. If the borrowed plural is not explicitly marked for gender (according to Arabic morphology) it has the same gender as the singular:

lbhim-t "domestic animal" (feminine), plural lbhaym (feminine)

lbzim "buckle" (masculine), plural lbzaym (masculine)

Loanwords whose singular is masculine may have a plural which is feminine, and marked as such (according to Arabic morphology), for example lɛlam "flag" (masculine), plural lɛlum-at (feminine).

The annexed state (EA) of an inflected noun is used in a number of clearly defined syntactical contexts:[35]

• when the noun occurs as subject in postverbal position:

tfulki it.is.beautiful tmazirt EA-country nnun of.you tfulki tmazirt nnun it.is.beautiful EA-country of.you "your country is beautiful"

tamazirt EL-country nnun of.you tfulki it.is.beautiful tamazirt nnun tfulki EL-country of.you it.is.beautiful "[as for] your country, it is beautiful"

• after most prepositions (see also § Prepositions):

tiskrt EL-garlic d with uẓalim EA-onions tiskrt d uẓalim EL-garlic with EA-onions "garlic and onions" (EL aẓalim)

tangult EL-loaf n of uɣrum EA-bread tangult n uɣrum EL-loaf of EA-bread "a loaf of bread" (EL aɣrum)

• after numerals 1 to 10 and after the indefinite numeral (see also § Numerals):

kraṭṭ three tmkilin EA-dishes kraṭṭ tmkilin three EA-dishes (EL timkilin)

mnnaw many wussan EA-days mnnaw wussan many EA-days (EL ussan)

• after some elements which require a following noun phrase (see also § Possessed nouns):

ayt Ugadir "the people of Agadir"

bu tɣanimt "he with EA-reed: flute player" (EL taɣanimt)

• after wan "like, such as" (premodern, obsolete in the modern language)

Outside these contexts, the EL is used. Uninflected nouns and unincorporated loans, which do not distinguish state, remain unchanged in these contexts.

The formation of feminine nouns from masculine nouns is a productive process. A feminine noun is formed by adding both the feminine nominal prefix t- (and, if necessary, a vocalic prefix), and the feminine singular suffix -t to a masculine noun. The semantic value of the feminine derivation is variable.

For many nouns referring to male and female humans or animals (mainly larger mammals), matching masculine and feminine forms exist with the same nominal stem, reflecting the sex of the referent:

adgal "widower" → tadgalt "widow"

amuslm "Muslim" → tamuslmt "Muslima"

ikni "twin boy" → tiknit "twin girl"

afullus "cock, rooster" → tafullust "hen"

izm "lion" → tizmt "lioness"

udad "moufflon" → tudatt "female moufflon"

In a few cases there are suppletive forms:

argaz "man, husband" ― tamɣart "woman, wife"

ankkur "buck" ― taɣaṭṭ "goat"

Feminine nouns derived from masculine nouns with inanimate reference have diminutive meaning:

aẓru "stone" → taẓrut "small stone"

ifri "cave" → tifrit "hole, lair"

lbit "room" → talbitt "small room"

ṣṣnduq "box" → taṣṣnduqt "little box"

urti "garden" → turtit "small garden"

Conversely, a masculine noun derived from a feminine noun has augmentative meaning:

tamda "lake" → amda "large lake"

tigmmi "house" → igmmi "large house"

tiznirt "fan palm" → iznir "large fan palm"

Feminine nouns derived from masculine collective nouns have singulative meaning:

asngar "maize" → tasngart "a cob"

ififl "peppers" → tififlt "a pepper"

bitljan "aubergines" → tabitljant "an aubergine"

luqid "matches" → taluqitt "a match"

Feminine derivations are also used as names of languages, professions and activities:

ahulandiy "Dutchman" → tahulandiyt "the Dutch language"

fransis "the French" → tafransist "the French language"

amzil "blacksmith" → tamzilt "blacksmith's profession"

inmmtri "beggar" → tinmmtrit "begging"

lmumsik "miser" → talmumsikt "avarice"

g°ma "(my) brother" → tag°mat "brotherhood"

There is an overlap here with feminine nouns denoting females:

tafransist "Frenchwoman" and "the French language"

tinmmtrit "beggarwoman" and "begging"

There are three deictic clitics which can follow a noun: proximal a-d "this, these", distal a-nn "that, those" (compare § Verbal deictic clitics) and anaphoric lli "the aforementioned":

tammnt ad ur tɣ°li "[as for] this honey, it is not expensive"

yaɣ usmmiḍ taɣaṭṭ ann bahra "the cold has badly afflicted that goat"

ifk ṭṭir lli i tazzanin ar srs ttlɛabn "then he gave the bird to some children to play with"

There are three basic sets of personal pronouns:

• independent
• direct object clitics
• suffixes

In addition, there are two derived sets which contain the suffixed pronouns (except in 1st singular):

• indirect object clitics
• possessive complements

Gender is consistently marked on 2nd singular, and on 2nd and 3rd plural. Gender is not consistently marked on 3rd singular and 1st plural. Gender is never marked on 1st singular.

The independent ("overt") pronouns are used to topicalize the subject or the object.

They are also used with certain pseudo-prepositions such as zund "like", abla "except":

The direct object clitics are used with transitive verbs:

The 3rd singular feminine variant stt is used after a dental stop, compare:

awi tt id "bring her here!" (imperative singular)

awyat stt id "bring her here!" (imperative plural masculine)

The direct object clitics are also used to indicate the subject with pseudo-verbs,[36] and with the presentative particle ha "here is, voici":

waḥdu yyi (alone me) "I alone"

kullu tn (all them) "they all, all of them"

laḥ t (absent him) "he's not there, he's disappeared"

manza tt (where her) "where is she?"

ha yyi (here.is me) "here I am"

The pronominal suffixes are used with prepositions to indicate the object (see § Prepositions), and with a closed set of necessarily possessed kinship terms to indicate possession (see § Possessed nouns). The plural forms add an infix -t- before the suffix with kinship terms, for example baba-t-nɣ "our father" (never *baba-nɣ); this infix also occurs with some prepositions as a free or dialectal variant of the form without the -t-:

flla-sn or flla-t-sn "on them"

dar-sn "with them" (never *dar-t-sn)

The indirect object clitics convey both benefactive and detrimental meaning:

The possessive complements follow the noun (see § Possession).

Prepositions can have up to three different forms, depending on the context in which they are used:

• before a noun or demonstrative pronoun
• with a pronominal suffix
• independent in relative clause

The form before nouns and demonstrative pronouns and the independent form are identical for most prepositions, the exception being the dative preposition i (independent mi, mu).

Most prepositions require a following inflected noun to be in the annexed state (EA) (see § Use of the annexed state). Exceptions are ar "until", s "toward" (in some modern dialects, and in premodern texts) and prepositions borrowed from Arabic (not in the table) such as bɛd "after" and qbl "before".

The instrumental and allative prepositions s "by means of" (with EA) and s "toward" (with EL) were still consistently kept apart in premodern manuscript texts. In most modern dialects they have been amalgamated, with both now requiring the EA, and with the pre-pronominal forms each occurring with both meanings: sr-s "toward it" (now also "with it"), is-s "with it" (now also "toward it").

The use of the different forms is illustrated here with the preposition ɣ "in":

Two prepositions can be combined:

Spatial relations are also expressed with phrases of the type "on top of":

ɣ iggi n umdduz "on top of the dung heap"

ɣ tama n uɣaras "beside the road"

ɣ tuẓẓumt n wasif "in the midst of the river"

The preposition gi(g)- "in" with pronominal suffixes, with all its free and dialectal variants,[37] is presented below. The other prepositions display a much smaller variety of forms.

The inherited cardinal numeral system consists of ten numerals (still in active use) and three numeral nouns (now obsolete) for "a tensome", "a hundred" and "a thousand". There is also an indefinite numeral meaning "several, many" or "how many?" which morphologically and syntactically patterns with the numerals 1 to 10. For numbers of 20 and over, Arabic numerals are commonly used.

These are listed below.[38] The formation of feminine "one" and "two" is irregular.

The numerals 1 to 10 are constructed with nouns (inflected nouns in the EA), the gender of the numeral agreeing with that of the noun:

yan "one wag°mar EA-horse" yan wag°mar "one EA-horse" yat "one tfunast EA-cow" yat tfunast "one EA-cow"

sin "two wag°marn EA-horses" sin wag°marn "two EA-horses" snat "two tfunasin EA-cows" snat tfunasin "two EA-cows"

[lower-alpha 5] The same obtains with the indefinite numeral:

mnnaw wag°marn "several/many EA-horses, how many horses?"

mnnawt tfunasin "several/many EA-cows, how many cows?"

Numerals yan, yat "one" also serve as indefinite article, for example yan urumiy "one Westerner, a Westerner", and they are used independently with the meaning "anyone" (yan), "anything" (yat):

ur iẓri ḥtta yan "he didn't see anyone"

ur ksuḍɣ yat "I'm not afraid of anything"

The final n of masculine yan "one" and sin "two" is often assimilated or fused to a following w, y or l:

yan w-ass → yaw wass "one EA-day"

yan w-sgg°as → ya wsgg°as "one EA-year"

yan lmakan → yal lmakan "a place"

sin y-sgg°as-n → si ysgg°asn "two EA-years"

sin y-ir-n → siy yirn "two EA-months"

The teens are made by connecting the numerals 1 to 9 to the numeral 10 with the preposition d "with". In the premodern language, both numerals took the gender of the counted noun, with the following noun in the plural (EA):

sin two d with mraw ten wag°marn EA-horses sin d mraw wag°marn two with ten EA-horses "twelve horses"

snat two d with mrawt ten tfunasin EA-cows snat d mrawt tfunasin two with ten EA-cows "twelve cows"

In the modern language, fused forms have developed in which the first numeral is always masculine,[39] while the following noun is in the singular, and connected with the preposition n "of":[40]

sindmraw twelve n of wag°mar EA-horse sindmraw n wag°mar twelve of EA-horse "twelve horses"

sindmrawt twelve n of tfunast EA-cow sindmrawt n tfunast twelve of EA-cow "twelve cows"

There are three inherited nouns to denote "a tensome", "a hundred" and "a thousand". These now seem to be obsolete, but they are well attested in the premodern manuscripts.[41] Morphologically, they are ordinary inflected nouns.

The tens, hundreds and thousand were formed by combining the numerals 1 to 10 with the numeral nouns:

snat two tmrawin EA-tensomes snat tmrawin two EA-tensomes "twenty"

snat two tmaḍ EA-hundreds snat tmaḍ two EA-hundreds "two hundred"

sin two wafḍan EA-thousands sin wafḍan two EA-thousands "two thousand"

The numeral nouns are connected with the preposition n "of" to a noun, which is most often in the singular:

timiḍi EL-hundred n of wag°mar EA-horse timiḍi n wag°mar EL-hundred of EA-horse "a hundred horses" snat two tmaḍ EA-hundreds n of wag°mar EA-horse snat tmaḍ n wag°mar two EA-hundreds of EA-horse "two hundred horses"

ifḍ EL-thousand n of tfunast EA-cow ifḍ n tfunast EL-thousand of EA-cow "a thousand cows" sin two wafḍan EA-thousands n of tfunast EA-cow sin wafḍan n tfunast two EA-thousands of EA-cow "two thousand cows"

In the modern language the Arabic tens are used, which have developed a separate feminine form:

ɛcrin twenty n of wag°mar EA-horse ɛcrin n wag°mar twenty of EA-horse "twenty horses"

ɛcrint twenty n of tfunast EA-cow ɛcrint n tfunast twenty of EA-cow "twenty cows"

The numerals between the tens are most frequently made with the Arabic numerals 1 to 10:

xmsa five w and ɛcrin twenty n of wag°mar EA-horse xmsa w ɛcrin n wag°mar five and twenty of EA-horse "twenty-five horses"

xmsa five w and ɛcrint twenty n of tfunast EA-cow xmsa w ɛcrint n tfunast five and twenty of EA-cow "twenty-five cows"

The Arabic hundreds and thousands are used in the modern language, taking the places of the original numeral nouns while the original syntax is maintained:

miya hundred n of wag°mar EA-horse miya n wag°mar hundred of EA-horse "a hundred horses" snat two id PL miya hundred n of wag°mar EA-horse snat id miya n wag°mar two PL hundred of EA-horse "two hundred horses"

alf thousand n of tfunast EA-cow alf n tfunast thousand of EA-cow "a thousand cows" sin two walfiwn EA-thousands n of tfunast EA-cow sin walfiwn n tfunast two EA-thousands of EA-cow "two thousand cows"

There is also a vigesimal system built on the Arabic numeral ɛcrin "twenty, score",[42] for example:

First and last are usually expressed with relative forms of the verbs izwur "to be first" and ggru "to be last":

tawriqqt page izwarn which.is.first tawriqqt izwarn page which.is.first "the first page"

ussan days gg°ranin which.are.last ussan gg°ranin days which.are.last "the last days"

There are also agent nouns derived from these verbs which are apposed to a noun or used independently:

aḍrf furrow amzwaru the.first.one aḍrf amzwaru furrow the.first.one "the first furrow"

tucka she.arrived d hither tamggarut the.last.one tucka d tamggarut she.arrived hither the.last.one "she arrived last"

The other ordinals are formed by prefixing masc. wis-, fem. tis- to a cardinal numeral,[43] which is then constructed with a plural noun in the usual manner:

wis-kraḍ ORD-three wussan EA-days wis-kraḍ wussan ORD-three EA-days "the third day"

tis-kraṭṭ ORD-three twal EA-times tis-kraṭṭ twal ORD-three EA-times "the third time"

The ordinal prefixes is also used with Arabic numerals and with the indefinite numeral:

wis-xmsa w-ɛcrin n dulqqiɛda "the 25th [day] of [the month] Dhū al-Qaʿda"

wis-mnnawt twal "the how-manieth time?"

Because four of the numerals 1 to 10 begin with s, the geminated ss that results from the prefixation of wis-, tis- (as in wissin, wissmmus, etc.) is often generalized to the other numerals: wissin, wisskraḍ, wisskkuẓ, etc.

A Shilha verb form is basically a combination of a person-number-gender (PNG) affix and a mood-aspect-negation (MAN) stem.

The workings of this system are illustrated here with the full conjugation of the verb fk "to give". The perfective negative goes with the negation wr "not". The imperfective goes with the preverbal particle ar (except usually the imperative, and the relative forms).

The verb fk "give" has the full complement of four different MAN stems:

• Aorist fk(i) ― fk in 1st, 2nd and 3rd singular, 1st plural, and the imperatives, but fki in 2nd and 3rd plural
• Perfective fki/a ― fki in 1st and 2nd singular, but fka with the other forms
• Perfective negative fki ― all forms
• Imperfective akka (an irregular formation) ― all forms

There are two basic sets of PNG affixes, one set marking the subject of ordinary verb forms, and another set marking the subject of imperatives.

Two suffixes (singular -n, plural -in) are added to the 3rd singular and masculine 3rd plural masculine verb forms respectively to make relative forms (also known as "participles"), as in i-fka-n "who gives", fka-n-in "who give".[lower-alpha 6]

A few verbs have just one MAN stem. The majority of verbs have two, three or four different MAN stems.[44] The Aorist stem serves as the citation form of a verb. The list below offers an overview of MAN stem paradigms. Around 15 paradigms of non-derived verbs can be recognized, based on the formation of the Perfective and the Perfective negative. Further subdivisions could be made on the basis of the formations of the Imperfective. All sections in the list contain a selection of verbs, except sections 12, 14, and 15, which contain a full listing.

The table below is adapted from Kossmann (2012:40, table 2.12 Uses of MAN stems in Figuig Berber).

Shilha has around twenty stative verbs which are still recognizable as a separate type of verb on the basis of their MAN stem paradigms. In earlier stages of the language, these verbs had their own separate set of PNG markers, which are sporadically found in premodern manuscripts:[45]

iḍ ɣzzif "the night, it is long" (cf. modern iḍ i-ɣzzif)

rẓag-t isafarn "medicines are bitter" (cf. modern rẓag-n isafarn)

In the modern language, these verbs take the regular PNG markers. Only the original singular relative form without prefix y- may still be encountered, for example adrar mqqur-n or adrar i-mqqur-n (mountain which.is.big) "big mountain". Stative verbs do not have a separate Perfective negative form. The table shows a selection of stative verbs.

There are two deictic clitics which are used with verbs to indicate movement toward or away from the point of reference: centripetal d "hither" and centrifugal nn "thither":

The use of these clitics is compulsory (idiomatic) with certain verbs. For example, the verb ack "come" almost always goes with the centripetal particle, and af "find" with the centrifugal clitic:

When the verbal deictic clitics occur after an object pronoun, they change to id and inn:

A possessive construction within a noun phrase is most frequently expressed as Possessee n Possessor. The preposition n "of" requires a following inflected noun to be in the annexed state. This kind of possessive construction covers a wide range of relationships, including both alienable and inalienable possession, and most of them not involving actual ownership:

anu n Dawd "Daoud's waterhole"

imi n tsraft "the entrance of the grain silo"

tarwa n Brahim "Brahim's children"

ig°dar n idqqi "pots of clay"

imikk n tisnt "a little salt"

atig n usngar "the price of maize"

tiɣ°rdin n imkli "after lunch"

lmdint n Ssnbul "the city of Istanbul"

aɣllay n tafukt "the rising of the sun"

aɣaras n ssk°ila "the road to school"

ddin n Wudayn "the religion of the Jews"

lqqiṣt n Yusf "the story of Joseph"

Many such possessive constructions are compounds, whose meaning cannot be deduced from the ordinary meaning of the nouns:

aɣaras n walim "road of straw: the Milky Way"

imi n wuccn "mouth of jackal: a length measure"[lower-alpha 9]

talat n tilkin "ravine of lice: nape, back of the neck"

tassmi n ifrgan "needle of hedges: kind of bird"

The possessor can itself be a possessee in a following possessive construction:

lmudda n tgldit n Mulay Lḥasan "the era of the reign of Moulay Lahcen"

luqt n warraw n wulli "the time of the giving birth of the sheep and goats"

As a rule, the preposition n assimilates to, or fuses with, a following w, y, l or m:[lower-alpha 10]

awal n w-aɛrab-n → awal w waɛrabn "the language of the Arabs"

a-ḍbib n y-isa-n → aḍbib y yisan "horse-doctor"

luq-t n w-nẓar → luqt unẓar "the season of rain"

a-gllid n y-muslm-n → agllid imuslmn "the king of the Muslims"

addag n litcin → addag l litcin "orange tree"

a-sngar n miṣr → asngar m Miṣr "maize of Egypt"

The possessor can also be expressed with a pronominal possessive complement. This consists of a pronominal suffix added to the preposition, which then takes the shape nn- (see § Pronouns). The form of the 1st singular possessive complement is anomalous: nw after a vowel, and inw after a consonant (or, in some dialects, niw):

agayyu nu "my head"

ifassn inu "my hands"

aḍar niw "my leg"

aqqrab nnk "your (sg.m.) pouch"

lumur nnm "your (sg.f.) affairs"

timlsa nns "his clothes"

rriy nns "her opinion"

aḍu nns "its smell"

adjarn nnɣ "our neighbours"

tawwuri nnun "your (pl.m.) occupation"

timddukkal nnunt "your (pl.f.) friends"

lmɛict nnsn "their (m.) livelihood"

tik°yaḍ nnsnt "their (f.) locks of hair"

There are two ways to express possession within a clause. The most common way is to use the "exist with" construction:

The verb ili "exist" (perfective lli/a) is usually omitted, leaving a verbless clause:

Alternatively, the verb ṭṭf, ṭṭaf "hold, possess" can be used:

In addition, there is the verb ili "possess" (perfective li/a), whose use is restricted to (inalienable) part-whole relationships and kinship relationships:

In al its usages ili can be replaced with ṭṭaf or the "exist with" construction, but not the other way around:

• azrg ila yan uskti, or azrg iṭṭaf yan uskti

(EL-hand-mill it.possesses one EA-handle)

"a hand-mill has one handle"

• ṭṭafɣ snat tg°mma, not *liɣ snat tg°mma

(I.possess two EA-houses)

"I have two houses"

These are a subtype of uninflected nouns. As with proper names, gender is not transparently marked on possessed nouns, which take gender agreement according to the natural sex of the referent. Plurals are either suppletive or made with the preposed pluralizer id. Most possessed nouns are consanguinal kinship terms which require a possessive suffix (the table contains a selection).

These kinship terms cannot occur without pronominal suffix. Example:

If these nouns are part of an NP-internal possessive construction, possession must be indicated twice:

The suffix must also be added when possession is expressed in a clause:

Some kinship terms are not possessed nouns but inflected nouns which take possessive complements (see examples above).

Another group of possessed nouns require a following noun phrase, occurring only in an NP-internal possessive phrase. A following inflected noun must be in the EA.

These four possessed nouns occur as first element in compound kinship terms (see above; w then becomes g° in g°-ma- "the brother of"). They also serve to indicate descent, origin and ethnicity:

Ḥmad u Musa "Ahmed son of Moussa" (name of a famous saint)[46]

u Brayyim "member of the Aït Brayyim ethnic group"

u bṛṛa "native of outside: a foreigner"

u Trudant "a native of Taroudant"

ayt Ugrsif "the natives of Aguercif"

ult Uglu "native woman of Aglou"

ist Tfrawt "the women of Tafraout"

When w is followed by another (phonemic) w the result is gg°:

w Wijjan → Gg°ijjan "native of Ouijjane" (also surname: Gouijjane)

a-rgaz w w-rgaz → argaz gg°rgaz "a man, son of a man: a man of virtue"

Ayt occurs in many Shilha ethnonyms:

Ayt Bubkr "the Sons of Boubker" (Aït Boubker), singular U Bubkr

Ayt Wafqqa "the Sons of Ouafka" (Aït Ouafka), singular w Wafqqa → Gg°afqqa

The proprietive elements masc. bu "he with, he of" and fem. mm "she with, she of" are borrowed from Arabic (original meaning "father of", "mother of"). They are used as formative elements and require a following inflected noun to be in the annexed state. The plural is formed with the pluralizer id:

bu sa yiwaliwn (he of seven EA-words) "a liar"

bu tagant (he of EA-forest) "wild boar"

bu tbratin (he with EA-letters) "postman"

id bu waga (PL he with EA-bucket) "French soldiers (wearing a képi)"

In many cases, bu fuses with a following nominal prefix:

bu w-marg (he with EA-poetry) → bumarg "a poet"

bu y-g°ra (he with EA-frogs) → Biyg°ra, Big°ra "Biougra" (place-name)

The feminine mm is encountered less frequently:

mm igrtal (she with EA-mats) "prayer room in a mosque"

mm lɣrur (she with delusion) "the world, worldly existence"

id mm ifalan (PL she with EA-threads) "needles"

The privative elements masc. war "he without" and fem. tar "she without" are made up of a gender prefix (masculine w-, feminine t-) and an element ar which is probably related to the negation wr "not". They do not require the annexed state, and should probably be translated as "who does not have", with the following noun phrase as object:

war acrik (he without EL-partner) "God"

id war tawwuri (PL he without EL-job) "the unemployed"

tar azal (she without EL-daylight) "wide-brimmed hat"

tar laman (she without certainty) "the world, worldly existence"