Phonology

1. Consonants
1.1. Overview

The following table shows the employed transcription letters that do not correspond to the IPA symbols.

The following chart lists short consonants that are common to all Senhaja varieties. 1 Consonants written in parentheses are rare or are in free variation with other phonemes. For example, the phoneme p and its long counterpart pp are mainly found in European loans, sometimes in free variation with b and bb,
and in some baby-talk lexemes. The phoneme ḍ [dˁ] is often in free variation with its spirantized counterpart ḏ̣ [ðˁ], and the phonemes ḷ [lˁ], ẓ̌ [ʒˁ], and <’> [ʔ] (the glottal stop, or the hamza), are limited to a handful of words. Senhaja ḍ/ḏ̣ can correspond to ṭ in Ghomara, e.g. Senhaja aḏ̣ar, Ghomara aṭar ‘foot, leg’; Senhaja aḏ̣i(l), Ghomara aṭil ‘grapes’. The phonemes č, čč are rare in Ketama, but are more frequent in other Senhaja varieties. Ketama šš often corresponds to čč in the rest of Senhaja, e.g. iḵeššem (K) vs. iḵeččem (most Snh.) ‘he enters’. The affricate ǧ is usually long (ǧǧ), but a few examples of a single ǧ also occur.
Short consonants common to Senhaja

Note: The following abbreviations are used in the charts: fric. = fricative, lab = labial, lbd = labialized, intrd = interdental, alv = alveolar, post-alv = post-alveolar, pal = palatal, vel = velar, uvu = uvular, phar = pharyngeal, lar = laryngeal; vd. = voiced, vl. = voiceless, fric. = fricative, phr. = pharyngealized, approx. = approximant.

Most consonants (except the glottal stop) can be long – in this case, they are doubled in the transcription. Spirantized consonants (written with a line below, e.g. ṯ [θ]) correspond to long stops, e.g. ṯṯ > tt. Exceptions are constituted by the long spirantized ḏ̣ḏ̣ [ðːˁ] that is in free variation with ḍḍ [dːˁ], and a rare
phoneme ḵḵ [çː] (that is both spirantized and long), as well as its labialized counterpart ḵḵʷ [çː w ]. Some short and long consonants are limited to specific Senhaja varieties. These are listed in the following chart.
Short and long consonants found dialectally

The spirantized ḇ is found in Taghzut, Hmed, Bunsar, and parts of Zerqet, while spirantized ḡ is limited to parts of Zerqet and Seddat. Labialized velars are found in most varieties, but have disappeared in Ketama and Taghzut. On tts and ddz in Zerqet, see further below.
1.2. Distinguishing features
Consonants across Berber varieties can be distinguished by voice, length, and pharyngealization.
1.2.1. Consonantal length
Pairs of short and long consonants are established in morphological oppositions, although in addition to length, other features can play a role. Short spirantized consonants correspond to long plosives, e.g. ṯ – tt, ḏ – dd, etc. As in other Berber varieties, we find the short/long pairs ḍ~ḏ̣ – ṭṭ, ġ – qq, and w – ggʷ.
For the last pair, w – ggʷ, we also find w – gg and w – ww correspondences in varieties lacking labialized velars (Ketama, Taghzut). In Zerqet, a short sibilant corresponds to an affricate long consonant: s – tts, ṣ – ṭṭṣ, z – ddz, š – čč, and ž – ǧǧ. The correspondence š – čč is also found in Seddat, Hmed, and Taghzut.
The following table shows correspondences of short sibilants in different Senhaja varieties.
Correspondences of short sibilants in Senhaja

The following table provides some examples of Aorist-Imperfective pairs in Zerqet.

The original *l – *ll pair has developed into different correspondences (cf. below): y – ll (Ketama/Hmed), ž – žž (Taghzut), y – ǧǧ (Seddat/Bunsar), l – ǧǧ
(Zerqet).
1.2.2. Pharyngealization
There are pharyngealized dentals and alveolars, some of which developed under the influence from Arabic. Pharyngealization is not always strongly pronounced in all words, and there are examples of depharyngealization. There are also lexemes that have developed pharyngealized phonemes, often as a result of the pharyngealization spread.
1.2.3. Spirantization
All Senhaja varieties underwent spirantization, which turned short stops into fricatives, e.g. t>ṯ, d>ḏ, k>ḵ. In most varieties, b is left unaffected, while in some varieties (Taghzut, Hmed, Bunsar, parts of Zerqet), b>ḇ. Similarly, in most varieties, g is left unaffected, while some varieties (e.g. parts of Zerqet and Seddat), g>ḡ. In some varieties, spirantized consonants have developed further into different sounds (stage 2 spirantization) or into a “zero”: t>ṯ >h (>Ø); g>ḡ>y (>Ø). The following table summarizes these developments.
Spirantization in Senhaja

Different from Tarifiyt (cf. Kossmann 2017), in Senhaja, k/g did not develop into š/ž. Hence, in some examples, Senhaja k and g correspond to Tarifiyt š and ž, respectively, e.g. Bunsar/Zerqet aggag, Hmed aggay vs. Tarifiyt ažžaž ‘thunder’; Ketama/Taghzut gir, Seddat ḡer, Zerqet ger/ḡer vs. Tarifiyt žar ‘between’.

Development ṯ > a-/h-/ah- (>Ø) in Ketama and Taghzut outside Lqela

In Ketama and parts of Taghzut (outside Lqela), ṯ > h (> Ø) in specific contexts. This change affects the 2S/3FS/2P verb subject prefix (ṯ- > h-/a-/ah-, but not a “zero”), and the feminine nominal prefix (ṯ > Ø). The Lqela dialect of Taghzut is conservative in preserving ṯ. Some examples follow.

Verb subject prefix (2S/3FS/2P): ṯ- > h-/a(h)-

Noun feminine prefix: ṯ >Ø

Development g > ḡ > y (> Ø)
The consonant g/ḡ has merged with y in some words in Eastern Senhaja (Hmed, Bunsar, Zerqet, Mezduy), whereas Western Senhaja (Ketama, Taghzut, Seddat) retains g. The following examples show this correspondence.

Correspondences of g ~ y in Senhaja

However, the correspondence is not always regular. There are lexemes where Ketama and Zerqet g corresponds to Hmed y (examples a below), and there are lexemes where g is preserved across Senhaja (examples b below):

Spirantization is often optional in many lexemes, e.g. kšem~ḵšem ‘to enter’, but there are also lexemes that prefer only one variant, either spirantized or not. Non-spirantized stops often originated from long consonants or borrowings. There can be dialectal differences in spirantization. Generally, Ketama (especially the dialect of Sahel) is less spirantizing than the Hmed variety.
Spirantization may be blocked in some cases, e.g. by the preceding n and m in Zerqet. This rule is not operative in Ketama or Hmed. For example:

The spirantization is not blocked by l or r, e.g.

On the whole, spirantization in Senhaja is not as regular as in Tarifiyt, and does not follow the same rules as Ghomara (Mourigh 2017).
1.2.4. Assibilation: t > ț
Assibilated ț [t͡s] is found in Taghzut, Hmed, and parts of Seddat; the phonemes t and tt are also encountered in these varieties. In Hmed, the assibilated ț can be followed by t or ṯ (without assimilation), e.g. nețṯa ‘she’. The exact conditioning of assibilation in these varieties differs. For example, the passive
prefix tt- can be (optionally) assibilated in Taghzut, but not in Hmed. In Hmed, ț- of the Imperfective prefix does not assimilate to the following t(t)- of the verb stem or the passive, e.g. țettežraḥ, Imperfective stem of the passive ttežraḥ ‘to
be injured’. In most other varieties (e.g. Ketama, Zerqet), the Imperfective and the passive t coalesce.

1.2.5. Labialized velars and uvulars
Labialized velars and uvulars exist in most Senhaja varieties, but are absent in Ketama and Taghzut. The existence of short labialized velars is an isogloss that distinguishes (parts of) Senhaja (excluding Ketama and Taghzut) from Tarifiyt (cf. Kossmann 2017: 97). Differently from Ghomara (Mourigh 2015: 59), the schwa adjacent to the labialized phoneme is not always realized as u in Senhaja. In some lexemes, labialization has been lost, and its traces are only visible in the vowel u. This is especially frequent in Ketama, but can be also found in varieties that preserve labialized velars. In Ketama, traces of lost
labialization are sometimes visible in the “unstable” u that corresponds to e in other Senhaja varieties and that disappears when the vowel occurs in an open syllable, e.g.
- suġ ‘buy! (SG)’ vs. sġ-aṯ ‘buy (PL)!’

The labialized phonemes in Senhaja are: ḵ(ḵ)ʷ, k(k)ʷ, ġʷ, g(g)ʷ. The phoneme ggʷ is the most frequent labialized phoneme in our database, largely because ggʷ is a long counterpart of w found in morphological oppositions. A single gʷ, by contrast, is rare, and has only been found in Hmed.
1.2.6. The fate of l and ll
The original *l became y (sometimes yy) in most Senhaja varieties (Ketama/Seddat/ Hmed, parts of Bunsar), ž(ž) in Taghzut, and is preserved in Zerqet. Rhotacism (l > r), as in Tarifiyt, is regularly found only in Mezduy. The following table shows some examples.
The original *l in Senhaja

In Ketama, in a few examples, l is in free variation with r, e.g. ala~ara ‘past marker’. In a few exceptional cases, l > r in Ketama, e.g. irket ‘louse’, cf. Zerqet ṯilk(k)iṯ ‘id.’. Word-final -el and -il became -i in varieties where l > y (i.e. -iy > -i),
e.g. aḏ̣il (Zerqet) > aḏ̣i (K/S/H/B) ‘grapes’.

The long *ll is preserved in Ketama, Hmed, as well as in Taghzut. However, a long counterpart of Taghzut ž (<*l) in morphological oppositions can be the regularized žž, e.g. mžeḵ (<*mleḵ) ‘to marry’, Imperfective mežžeḵ. Usually, long ll > ǧǧ in Seddat, Bunsar, Zerqet, and Mezduy (with some exceptions), as in Tarifiyt. The following table shows some examples.
The original *ll in Senhaja

The correspondence *l –*ll in the Aorist-Imperfective pairs

8
There are some exceptions: l does not always become y in the varieties where this change took place, while ll does not always become ǧǧ in the varieties where this change took place, especially in Arabic loans. In feminine nouns that have the suffix -ṯ, the final -lṯ may be preserved, or l may undergo the usual
changes. The final -lṯ changes to -šṯ in some dialects of Zerqet, parts of Seddat and Bunsar, e.g. ṯagfilṯ > ṯagfišṯ ‘egg’. In parts of Seddat, Bunsar, and Mezduy, -lṯ > č [tš], e.g. ṯagfilṯ > ṯagfič ‘egg’.

2. Vowels
Senhaja has three peripheral vowels: a, usually realized as [æ], i [ɪ], and u [u], and a central vowel e [ə] (schwa).
Senhaja vowel system

When any of the peripheral vowels is adjacent to a pharyngealized consonant, it is backed and lowered. Schwa is realized as [u] before w, and as [i] before y. Word-finally, -ew > u and -ey > i (including <*-el).
In the final syllable, before back consonants (uvulars, pharyngeals, and laryngeals), the difference between schwa and a is neutralized in most Senhaja varieties (both are realized as [a]), with the exception of Taghzut (Lqela), parts of Seddat and Zerqet, and Mezduy, e.g. žreḥ (T/M/parts of S/Z), žraḥ (the rest of Snh.: K/H/B/most Z) ‘to injure’.
In Senhaja, as elsewhere, the presence and position of schwa in a word can usually be predicted by means of an insertion rule. Schwa is generally inserted from right to left between two consonants, schematically: (C)CC > (C)CeC (‘structure-based syllabification’ in Kossmann 2012a). When due to the addition
of affixes or clitics, schwa comes to stand in an open syllable, schwa is elided or the word is resyllabified, e.g.
- ḵšem! ‘Enter! (sg.)
- ḵešm-en ‘they entered’.
There are exceptions from this general rule, e.g. words ending in CeCC instead of the expected CCeC, e.g. izerf (S/B/Z) ‘road’, cf. izref (K/T/H) ‘id.’. Loans from Moroccan Arabic also often present exceptions. We conclude, following Saïb 1976a and Kossmann 1995a, that schwa is mostly predictable (inserted following a set of rules), but sometimes phonemic (inherent).
There is a contrast between CyeC# vs. CiC#, and between CweC# vs. CuC#.
The semivowels y and w are distinguished from the high vowels i and u, but in certain contexts, y may become i, and w may become u. Etymological word-final -ew became -u, and word-final -ey became -i.
3. Consonant Assimilations and Vocalic Sandhi

Consonant assimilations can occur within a word or across word boundaries. Vocalic sandhi occurs across word boundaries and involves vowel elision, change of a vowel into a semivowel, or Hiatustilger (insertion of a semivowel to break up the sequence of two vowels, or hiatus). In our work, the symbol ^indicates vowel elision, truncation, and assimilation across morpheme
boundaries.
3.1. Consonant assimilations
Consonant assimilations are normally regressive, with the exception of the preverbal ventive clitic d. Consonants that frequently undergo assimilation are alveolar stops and post-alveolar fricatives. Assimilations in voice are very common: voiceless + voiced > voiced (e.g. t + d > d^d), and voiced + voiceless > voiceless (e.g. d + t > t^t, ġ+ṯ > ḫ^ṯ). Not all assimilations take place in all dialects. Generally, fewer assimilations are observed in Hmed (especially in the
verbal complex) than in Ketama and Zerqet.
Sibilants often assimilate to each other, including long-distance assimilation (sibilant harmony), e.g. s+š > š^š. The consonant n often assimilates to the following l (n+l > l^l), and can also assimilate to the following ḷ, r, and ṛ, but does not assimilate to other consonants. Assimilation of n affects the 1P subject
prefix n-, the Genitive preposition n ‘of’, and the final -n of the numeral ‘one’, among others.
3.2. Vocalic Sandhi
A sequence of two vowels is normally not permitted in Berber.  When two vowels come in contact, the following solutions are possible:
1) one of the vowels is elided (vowel elision), e.g. a + a > a, e.g.
i-nna + as > i-nna^s ‘He said to him/her’
2) a semivowel is inserted (Hiatustilger), e.g. a + a > a ya, e.g.
a + arba > a yarba! ‘o boy!’;
3) one of the vowels becomes a semivowel, e.g. a + i > a y, e.g.
arba + inu > arba ynu ‘my son’.
4. Summary of isoglosses in phonetics/phonology
The following table summarizes the isoglosses in phonetics/phonology within Senhaja. Comparisons with Ghomara (Ghm.) and (Central) Tarifiyt (Rif) are provided.
Summary of isoglosses in phonetics/phonology