Emphatic consonants beyond Arabic: The emergence and proliferation of uvular-pharyngeal emphasis in Kumzari

3.1 Emphatic consonant inventories

Phonologically speaking, the term “emphasis” is prototypically applied to a series of dental consonants, well-known from Classical Arabic, which are distinguished from their “plain” counterparts through a salient co-articulation (Ar. ʔiṭbāq ‘spreading and raising’) most commonly defined as pharyngealization (Card 1983: 13–14; Watson 2002; Bakalla 2009: 421) or velarization (Fischer and Jastrow 1980: 56; Holes 2004: 57; Embarki 2013), but also treated as uvularization by some scholars (Dolgopolsky 1977: 1; McCarthy 1994; Shahin 1997, Shahin 1998; Zawaydeh 1998). There are four contrastive emphatics in Classical Arabic (CA) and, by extension, Modern Standard Arabic (MSA) which fall under this label: ṭ [t̪ˤ], ḍ (CA [ɮˤ], MSA [d̪ˤ]; see Section ‎5.4.2), ṣ [s̪ˤ], and đ̣ [ðˤ] (Bakallah 2009: 421; Lehn 1963: 29).

The idea of emphasis is generalized under the Arabic label tafxīm (lit. ‘thickening, enlarging, emphasizing’) to additionally designate emphatic allophonic variants of the consonants l and r in CA and MSA (Owens 2006: 25; Holes 2004: 57–58; cf. Ferguson 1956) as well as similarly co-articulated consonants in other dialects (Jakobson 1957/1971; Bakallah 2009: 421; Grigore 2011). In some cases, inventories of contrastively emphatic consonants are larger than that of Classical Arabic: for example, Nigerian Arabic has six underlyingly emphatic consonants ṭ ḍ ṣ ṃ ḷ and ṛ, as well as emphatic allophones of several additional consonants (Owens 2006: 25); and the Northeastern Neo-Aramaic dialect of Amedia in Iraq contains ten contrastive pharyngealized consonants: p̣ ḅ ṭ ḍ č̣ ṣ ẓ ṃ ḷ and ṛ (Hoberman 1985: 224).

In a seminal article on the linguistic nature and behaviour of emphasis, Lehn (1963: 29) claimed (without providing a reference) that writers in the medieval Arabic grammatical tradition included the characteristically Semitic pharyngeal consonants ḥ (typically [ħ] or [ʜ]) and ʕ ([ʕ], [ʢ], [ʡ]) ^[3] with the emphatic dentals under the label of tafxīm. This assertion, for which I have not been able to locate an original source, has been taken up and explored by other linguists (Paddock 1970 in Card 1983: 91; see also Marçais 1948). In such a framework, ḥ and ʕ would be viewed as the emphatic counterparts of the glottal fricative h and the glottal stop ʔ (cf. Jakobson 1957/1971: 519; Hoberman 1985: 227–228; Kahn 1976: 28–29; Watson 2002: 268). This assessment recognizes a common post-velar articulation for co-articulated emphatics and pharyngeals, and some similar phonetic effects of both groups (see Sections ‎3.2 and ‎3.4 below).

In the work of the earliest Arabic grammarians and in much of the literature today, however, ḥ and ʕ are excluded (Lehn 1963: 29). This arises from the fact that, in Arabic, there is little diachronic connection between co-articulated emphatics and pharyngeals (Hoberman 1985: 223; cf. Section ‎3.4 below). In most varieties of Arabic, they are also synchronically discrete; and in Cairo Arabic and Palestinian Arabic, there is even a contrast available between plain and emphatic pharyngeals (Lehn 1963: 31–32; Card 1983: 18, 22). From the perspective of current phonological theory, this may seem unusual, since pharyngeals and pharyngealized consonants are often grouped together as a natural class with a shared distinctive feature of [pharyngeal], [RTR] (retracted tongue root), [guttural] or something similar (Jakobson 1957/1971; Clements and Hume 1995: 273–274; Hayward and Hayward 1989; cf. Davis 1995: 472). However, as Card (1983: 16) and Davis (1995: 483) have pointed out, there are phonetic as well as phonological differences between the two groups of consonants in Arabic (see also Sections ‎3.2 and ‎3.3 below).

Despite these differences, and in light of the great variation in patterning across languages, it is worth reconsidering the idea that in some situations, pharyngeal consonants can in fact be classed along with co-articulated emphatics as part of a broader meaning of emphasis. Examples of historical interaction between co-articulated emphatics and pharyngeal consonants have been observed in sporadically distributed varieties of Arabic (Brockelmann 1908 and Blanc 1953 in Hoberman 1985: 223), but no consistent patterns of interaction have emerged. Some synchronic support for a relationship between the two groups of consonants is, however, found in certain varieties of Kurdish: there, the two groups not only demonstrate similar phonetic behaviour (e. g., phonetic effects on vowels; see Kahn 1976: 22), but also exhibit complementary distribution in some contexts (Kahn 1976: 87–89) and free (or stylistic) variation in others (e. g., taʕzi ~ ṭazi ‘fresh’) (Kahn 1976: 50; Hoberman 1985: 229). Further evidence that the two groups could constitute a single natural class comes from the observation that in some languages, pharyngeal consonants are consistently correlated with the diachronic retention or even appearance of contrastive emphasis on other potentially emphatic consonants; this will be explored in ‎3.4 below.

3.2 Articulatory properties of emphasis

As mentioned in the preceding section, the articulatory basis of emphasis is usually identified variously as pharyngealization, velarization, or uvularization. In fact, however, it is almost always (or perhaps always) the case that emphasis functions as a complex of secondary articulations which can also include labialization, glottalization and other co-articulations (Jakobson 1957/1971: 511–513; Hoberman 1988, Hoberman 1989: 77, Hoberman 1995: 841; Watson 2002: 269; Embarki 2013: 31–34). This complexity is reflected in an abundance of labels in the linguistic literature, as listed by Lehn (1963: 29): along with pharyngealization, velarization and uvularization, the terms retraction, strong articulation, u-resonance and heaviness have been applied to emphasis.

In articulatory terms, emphatic consonants are typically characterized at least by a narrowing of the upper pharynx (Watson 2002: 269). Acoustically, this narrowing results in the lowering of the second formant (Card 1983: 13–15; cf. Jakobson 1957/1971: 512; Watson 2002: 270). Emphatic consonants also have in common an effect on neighbouring vowels, in particular the low vowels a and ā (Bakallah 2009: 422; see Section ‎3.4 below). Hoberman (1989: 77) questions the idea that the production of emphasis necessarily varies from one variety to another, but a wide range of articulatory analyses have been proposed for emphasis in different languages and dialects. In Card’s (1983) study of Palestinian Arabic, pharyngealization is equated with emphasis; but Hoberman (1985: 224) observes that pharyngealization is an insignificant component of emphasis in Iranian dialects of Neo-Aramaic. Elsewhere, in varieties such as Shihhi Arabic and Kumzari, stricture at a higher location, namely uvularization, is dominant (cf. Anonby 2011a: 376).

An important point introduced in the previous section, and one that recurs in the literature on emphatics, is the distinction between the secondary articulation of emphatics and the primary articulation of pharyngeal consonants ḥ and ʕ. Although Card (1983) highlights some similarities between the two sets of sounds in Arabic, she concludes that their articulation is fundamentally different: whereas the main secondary constriction of co-articulated emphatics is located in the upper pharynx, the primary constriction of pharyngeals is located consistently further down in the pharynx; and that whereas acoustically, co-articulated emphatics exhibit a strong lowering of the second formant, the effect of pharyngeals on the second formant should be seen as incidental (pp. 13–15, 90–97; this inference will, however, be revisited in the next section). It is precisely this state of affairs which motivates Dolgopolsky (1977: 1), followed by Shahin (1997) and Zawaydeh (1998), to argue that Arabic emphasis should be referred to as uvularization rather than pharyngealization; in Dolgopolsky’s words, this would distinguish it from “lower” pharyngealization such as that found in languages of the Caucasus (see also Colarusso 1975).

3.3 Phonological parameters of emphasis

In the simplest scenario, such as that which is often put forward for Classical Arabic (CA) and Modern Standard Arabic (MSA), emphasis may be viewed as an intrinsic phonological property of emphatic consonants. While recognizing that emphatic consonants have predictable phonetic effects on neighbouring segments, most accounts of the phenomenon in CA and MSA, and even some accounts of spoken Arabic varieties, have taken for granted the idea that emphasis can only have as its phonological domain the emphatic consonants themselves (cf. the publications listed in Lehn 1963: 34–35).

In most spoken varieties of Arabic, and in other languages with emphasis, the segment is not in fact the minimal domain; the domain with which this feature is associated tends to vary from one dialect to another (Watson 2002; Holes 2004: 58). In Cairo Arabic, for example, the minimal underlying domain of emphasis is the syllable (Lehn 1963: 32; Hoberman 1989: 83); and in Qatari Arabic (Bukshaisha 1985: 217–219), as well as Neo-Aramaic varieties spoken in Iran, it is typically associated with whole words (Hoberman 1985: 225). In Neo-Aramaic in particular, there are many words in which no intrinsically emphatic underlying consonant can be distinguished.

In most varieties, though, the ultimate source of emphasis is in fact situated in underlyingly emphatic consonants, also known as “primary” emphatics (Fischer and Jastrow 1980: 56–57). Yet often, even here it is not confined to the emphatic consonants themselves; rather, it spreads synchronically within a word according to specific structural constraints (Watson 1999, 2002). In their work on phonological parameters of emphasis, Hoberman (1989: 89–90) initiates and Davis (1995) and Watson (1999, 2002) further develop an overview of three parameters which are particularly helpful in understanding the source and synchronic spread of emphasis in such languages: direction of spread, transparent consonants, and opaque consonants. Regarding the direction of spread in particular, there are variations in different languages and dialects: whereas in Palestinian Arabic, emphasis spreads both progressively (to the right) and regressively (to the left) within a word (cf. also Davis 1995), it spreads only progressively in the Christian Neo-Aramaic dialect of Iranian Azerbaijan; in Kurdish, emphasis does not spread, being systematically limited to a single consonant in any word (Kahn 1976: 49). The extent of spreading in languages where it does occur is affected by the presence of two kinds of segments: “transparent” segments, which although they cannot themselves be associated with the emphatic feature, neither block nor initiate spreading; and “opaque” segments which block the spread of emphasis because they are invariably specified as not associating with emphasis. Examples of transparent segments are w and ū in Palestinian Arabic; and in several varieties, the high, non-back segments y, ī and i are opaque (Fischer and Jastrow 1980: 57; Card 1983: 15, 118). Hoberman (1985: 224) suggests further that many other dialects of Arabic make use of opaque and transparent segments, as do Neo-Aramaic dialects outside of Iran.

3.4 Variations in the description of the scope of emphasis

The parameters outlined in the preceding section provide a satisfactory account of the way in which emphasis functions synchronically, but the scope of the phenomenon differs significantly in the various accounts. The limited data available on the topic suggest that diachronic sound changes involving emphasis may operate by means of a more general, or less clearly-defined, mechanism.

There are a number of additional consonants, most of which are post-velar, which in various frameworks have been recurrently identified with co-articulated emphatics: r w x ġ q ḥ and ʕ. Importantly, along with the core emphatic series ṭ ḍ ṣ đ̣, which was characterized by ʔiṭbāq (see ‎3.1 above), the early Arabic grammarian Sibawaih, and those following his work, considered x ġ q (but not w) a second subset of consonants exhibiting ʔistiʕlāʔ ‘raising (of the tongue back)’ (Sibawaih II.452 [1970], Bakallah 2009: 421; cf. Jakobson 1957/1971: 515–516), although his justification for the broader category was related to phonological effect on neighbouring vowels rather than the mechanics of production. ^[4] As mentioned above (Section ‎3.1), Lehn claims that later Arabic grammarians included the pharyngeals ḥ and ʕ as part of this larger phonological set of ʔistiʕlāʔ (Lehn 1963: 29), but since no reference to his assertion can be traced, this point remains in doubt. In the contemporary literature, Holes (2004: 57–58) observes an articulatory parallel between co-articulated emphatics and the consonants r x ġ q in that all of them result in the backing of neighbouring vowels. While Card (1983: 90–97) dismisses a phonological relationship between r w ḥ ʕ and co-articulated emphatics, she notes that the acoustic lowering of the second formant characteristic of co-articulated emphatics is to some degree present in these consonants, especially r and w. Finally, McCarthy (1994: 218–223) and Herzallah (1990: 55) argue for a class of [pharyngeal] “gutturals”; this class, based primarily on co-occurrence patterns, includes the co-articulated emphatics, the uvulars x ġ q, and the pharyngeals ḥ and ʕ; additionally, in contrast to many authors, they include the laryngeals h and ʔ in this set.

Although the consonants r w x ġ q are not themselves contrastively emphasized, and although (as discussed above in ‎3.2) there are important phonetic and phonological differences between co-articulated emphatics and pharyngeals ḥ and ʕ in many varieties, there is in certain languages a strong correlation between some of these consonants and the diachronic maintenance of emphasis, or even its appearance on neighbouring, potentially emphatic consonants (i. e., non-emphatic consonants for which an emphatic counterpart exists in the language). Hoberman (1985) carefully demonstrates that in two dialects of Northeastern Neo-Aramaic (the Christian dialect of Urmi and the Jewish dialect of Iranian Azerbaijan), words which in Old Aramaic contained either pharyngealized consonants or the pharyngeals ḥ and ʕ are reflected as emphatic words (that is, entire words which carry emphasis; see Section ‎3.3). Similarly, in the Jewish Neo-Aramaic variety of Zakho, the two pharyngeals as well as q have been retained in words which in Old Aramaic also had co-articulated emphatics, even though they are dropped in other environments (pp. 224, 228–229). In some Kurdish dialects, where emphasis can be marked on only one point within a word, other emphatic consonants cannot coexist with q; this suggests that it is the emphatic counterpart of k (Kahn 1976: 23, 26–27, 87–88). In these same dialects, a borrowed ʕ has been reinterpreted as pharyngealization on a consonant elsewhere in the word (e. g., Ar. zaʕfaran>Kurdish ẓafaran ‘saffron’; p. 50). Historical and comparative examples are limited for r, but in the Qabbe dialect of Shihhi Arabic, where the consonant r is strongly uvularized, the most likely explanation for the appearance of emphasis on the final consonant of rōṣ ‘head’(cf. Classical/Modern Standard Arabic ra’s) is the co-occurrence of r (author’s field notes, 2010). Critically, as will be shown for Kumzari, the presence of w x q and ḥ have repeatedly led to the diachronic appearance of emphasis on potentially emphatic consonants within a word, according to the same mechanism by which emphasis diffuses from co-articulated emphatics (Section ‎5.3.1). Conversely, for ḥ in particular, there is evidence that this pharyngeal may arise as a result of emphasis spreading onto h from the Kumzari emphasis-inducing consonants mentioned here (Section ‎5.3.2).

In short, it is clear that the functioning of emphasis, and the boundaries of such a natural class, vary across languages and can be difficult to define in any given language. The manifold possibilities in the evolution and patterning of emphasis are likely traceable to the phonetic and phonological complexity of the phenomenon itself (Section ‎3.2). As regards diachronic maintenance and spread of emphasis, the consonants r w x and q function in some languages like co-articulated emphatics; in this sense, they may be regarded within these systems as somehow intrinsically emphatic. In a discussion of emphasis, then, the idea of opacity (i. e., the obligatory specification of a segment with a feature) need not always be limited to consonants that block the synchronic spread of emphasis (cf. Section ‎3.3 above); it should be extended to intrinsically emphatic segments. As for ḥ and ʕ, the synchronic evidence that pharyngeals could in some languages be considered emphatic counterparts of h and ʔ (Section ‎3.1) is strengthened by the parallels between the two pairs: centrally, the fact that both can function in the same way as co-articulated emphatics with regard to the diachronic maintenance, or even induction, of emphasis on neighbouring consonants; and, as shown in Kumzari, the contrastively emphatic nature of ḥ as reflected by the recasting of h as ḥ through emphatic diffusion (Sections 5.2.2 and ‎5.3.2).

3.5 Emphasis in language contact situations

Emphasis in language contact situations is consistently characterized by instability. In areas where Central Semitic and languages from other families are in contact with each other, major changes in its status and distribution have taken place in both directions.

To begin with, languages which historically contained emphatic consonants have responded in two ways. In most varieties of Arabic, emphatic consonants have not only been maintained in contact situations; in many cases, non-emphatic consonants in Arabic words borrowed from contact languages have been reinterpreted as emphatics: for example, ʔaṭlas ‘atlas’ (from Greek átlas), baṭāṭā ‘potato’ (from Portugeuse batata), ṣard ‘extreme cold’ and ṭāzaj ‘fresh’ (from Middle Persian sard ‘cold’ and tāzag ‘flowing, fresh’). (The phonetic basis of this type of reinterpretation is explored in Jakobson 1957/1971: 512–513.) This contrasts with the situation of several geographically peripheral varieties of Arabic, including Maltese, some dialects of Chadian and Sudanese Arabic, and Arabic spoken in Uzbekistan, which as a result of intense contact with other languages have lost the emphatic series from their consonant inventories (Fischer and Jastrow 1980: 56).

Conversely, varieties in contact with languages containing emphatic consonants have responded in a number of ways. In most cases, even when contact varieties have undergone extensive lexification from Arabic, they conform borrowed words to their own phonological inventories, in particular with respect to emphatic consonants (Paradis and LaCharité 2001). Persian and Turkish are two important examples of this. In other cases, the phonological impact of contact is more substantial, and a subset of emphatic consonants has been retained in borrowed words; this is the situation for Keshmi (Section ‎4.4), Pashto (Section ‎4.5) and Swahili (Lodhi 2003: 157), although in the case of Pashto and Swahili (and, most likely, many other languages in the broader Islamic sphere), emphatic consonants are further restricted to formal acrolects. Finally, there are a limited number of languages in which emphasis, having arisen through lexical borrowing, has been integrated into the phonological system as a result of intense, long-standing contact. ^[5] In these languages, emphasis is not limited to borrowed vocabulary, but over time begins to appear – to varying degrees – in inherited (i. e., non-borrowed) vocabulary. Examples of languages in which emphasis has become integrated to such an extent are Kurdish (Section ‎4.1), Zazaki (Section ‎4.2), Dezfuli-Shushtari (Section ‎4.3), Domari (Matras 2007: 152), some Berber languages (Applegate 1970: 592–593, 604) and, as will be explored in depth below, Kumzari (Sections ‎4.6–‎5.5). In each situation, and often even for subdialects of the same language, the mechanisms driving the proliferation of emphasis appear to be unique (Hoberman 1989: 91), but there are no comprehensive accounts of how this happens in any particular variety. The present paper has been written in response to this question.

4.1 Kurdish

The best-known case of emphasis in the phonological inventory of an Iranian language is that of the (non-Southwestern) West Iranian language Kurdish; but even here, accounts of the phenomenon are for the most part fragmentary.

MacKenzie (1961) observes the existence of pharyngeal and co-articulated emphatic consonants in a number of Kurdish dialects. The most significant contribution to the topic is that of Kahn (1976), who points out that the behaviour and extent of emphasis in Kurdish varies widely among dialects, and even among speakers of the same dialect (25, 106–108). Kahn examines emphatic consonants as part of a description of borrowing and variation in Northern Kurdish (“Kurmanji”) dialects spoken around Rezaiyeh in the West Azerbaijan Province of Iran. There, emphasis is contrastively (but variably) found with a range of consonants (see the discussion below), yet shows a restricted distribution in the lexicon as well as the phonology (43–44). Most occurrences of emphatic consonants in these dialects are simple retentions in Arabic loanwords; emphasis has, however, arisen spontaneously in a limited number of inherited vocabulary items, and Arabic (but not Persian or Turkish) vocabulary is also susceptible to reinterpretation (see the examples in this section below). While the presence of low vowels æ [æ] and a [ɑ] or the back vowel o [o] does not necessitate the appearance of emphasis, it is a recurring phonological context in which emphasis arises on consonants (90). Due to a language-specific constraint whereby emphasis can be associated with only one consonant in a word, it never spreads, either diachronically or synchronically, from one emphatic consonant to another (42–51) (although in borrowed words it is occasionally displaced from one consonant to another to provide a more favourable location, generally closer to the beginning of a word; e. g., Ar. zaʕfarān>Kurdish ẓæfæran ‘saffron’, p. 50).

Surprisingly, it is in the Northern dialects of Kurdish – which have had less direct contact with Arabic – that the phenomenon appears to be most deeply established (28). The question of whether the initial emergence of emphasis can therefore be at least partially attributed to longstanding contact with Aramaic (cf. Section ‎3.5) has been raised (cf. Windfuhr’s comment in Hoberman 1985: 229; Haig 2007: 165; cf. Kahn 1976: 5, 12), but not addressed.

In any case, as mentioned above, most Kurdish words with emphatic consonants are borrowed from Arabic. It is widely recognized that Kurdish retains the pharyngeal consonants ḥ and ʕ in Arabic loanwords (Northern Kurdish data from Kahn 1976: 88):

In some dialects, the inventory of borrowed emphatics extends to the pharyngealized alveolars ṭ ṣ and ẓ (28, 81, 88).

A survey of the literature reveals, however, that in various dialects there are a number of cases where these phonemes have diffused into inherited Iranian vocabulary. The Kurdish examples below are taken from Kahn (1976: 25–29, 50 and 108) unless otherwise indicated.

Alongside ḥ ʕ ṭ ṣ and ẓ, there are dispersed accounts of additional contrastive emphatics in inherited vocabulary. The best-known of these is the velarized or pharyngealized alveolar lateral approximant ḷ ([ɫ]=[lˠ]/[lˤ]), which contrasts with a plain alveolar counterpart l in Central and Southern Kurdish varieties (Haig and Matras 2002: 5; Kahn 1976: 32). This opposition is confirmed by minimal pairs such as guḷ ‘rose’ and gul ‘person with leprosy’ (McCarus 2009: 592). Additional pairs from Central Kurdish are kæḷ ‘water buffalo’ vs. kæl ‘hill’, pæḷæ ‘spot, stain vs. pælæ ‘rush (n.),’ and kuḷ ‘sorrow’ vs. kul ‘dull’ (Jaffer Sheyholislami, pers. comm. 2012).

In her overview of Kurdish, Blau (1989b: 329) mentions that Kurdish varieties spoken in Armenia also contain a contrastive voiceless pharyngealized palato-alveolar affricate č̣ [t͡ʃ ˤ], and Kahn (1976: 27–28) provides the examples č̣æng ‘fistful’ and peč̣an ‘to wrap’ from Northern Kurdish in Iran.

Kahn cites the additional occurrence of a contrastive p̣ [pˤ] in Northern Kurdish (25–26):

A further consonant in this set, ṛ, can be realized as either flapped [ɾˤ] or trilled [rˤ] but for which emphatic co-articulation is distinctive. It is found in some varieties of the Mokriyani dialect of Central Kurdish in place of the (reportedly) non-emphatic trilled phoneme ř known from other Kurdish varieties (data from Jaffer Sheyholislami, pers. comm. 2012).

As mentioned in ‎3.4 above, Kahn (1976: 26–27) includes the voiceless uvular stop q as a member of the Kurdish emphatic series, since its distribution mirrors that of other emphatic consonants.

Finally, as stated earlier in this section, it should be pointed out that there are number of Arabic loanwords in spoken Kurdish where the emphatic is not present in the source item, but has arisen in Kurdish subsequent to borrowing (Kurdish items from Kahn 1976: 89–90):

Specifically, Kahn sees the introduction of emphasis on non-emphatic consonants in Arabic borrowings as a type of hypercorrection which is at least partly conscious, and labels it “hyper-Arabization” (Kahn 1976: 89–90). This contrasts with the situation in written Kurdish, where Arabic features such as emphasis are deliberately underrepresented. Barry (2019) prefers an analogical, phonetic explanation for the appearance of emphasis in inherited Northern Kurdish structures and, remarkably, shows a strong but imperfect correlation between the emergence of ḥ and ʕ (which he reanalyzes as a pharyngealized vowel feature) and the presence of labials.

Taken together, these data show that emphatic consonants have been phonologized in Kurdish in various ways, and to various degrees. Such consonants are found in most Kurdish varieties, but in each case, they are relatively uncommon and appear to be restricted to the periphery of the grammar. Haig’s observations concerning the Central Anatolian dialects of Northern Kurdish appropriately sum up most scholars’ view of the situation for the language as a whole: “In a sense, the pharyngeals are extraneous to the basic phonology: they are restricted to individual lexical items, they play no part in morphology, their functional load is very limited, and there is considerable cross-speaker and cross-dialect variability in the extent of their presence” (2007: 167). However, Barry’s (forthcoming) findings leave open the possibility of a system in which, as in Kumzari, emphasis is becoming increasingly phonologized and, in fact, may have a connection to the pharyngeal segments traditionally analyzed as phonemes ḥ and ʕ.

4.2 Zazaki

Little has been written on the status of emphatics in Zazaki, a (non-Southwestern) West Iranian language spoken in the north-western portion of the larger Kurdish language area but fundamentally distinct from Kurdish (Paul 2009: 545).

Some of the properties of emphasis in Zazaki appear to be the same as those of Kurdish, although this topic is not addressed in detail in the literature. In their overviews of Zazaki, Blau (1989a: 338) mentions a dental emphatic series, and Paul (1998: 3–6, 11, 2009: 547) presents in the consonant inventory several contrastive but infrequently occurring consonants which could be labelled emphatic: ṭ, ṣ, velarized ḷ, the pharyngeal consonant ḥ and, in some varieties, the pharyngeal consonant ʕ.

While examples of most of the emphatics are not listed along with the inventory, a number of Zazaki words distributed throughout Paul’s studies contain the pharyngeal consonant ḥ in particular. As is the case for this consonant in Kurdish, most of the words in which it is found are borrowings of Arabic items that contain ḥ:

In a few instances, however, Paul shows that Zazaki ḥ is a reflex of Arabic ʕ (e. g., bāḥd / bāḥdo<Ar. baʕd ‘after’), and for the numeral ‘seven’, in parallel with Kurdish, an inherited Iranian root haft ‘seven’ is rendered with an emphatic: ḥewt.

The emphatic ʕ, in turn, is occasionally carried over from Arabic (ʕerd<Ar. ʔarḍ ‘ground’), and in a few cases replaces the glottal stop ʔ in words of Arabic origin (e. g., ʕemel<Ar. ʔamal ‘hope’). The emphatic ṣ is found in one Iranian root (ṣī ‘stone’), and ḷ is attested in the item ḷīmin ‘dirty’; and the emphatic ṭ is only found in words of Arabic origin, such as ṭeyr (>Ar. tayr ‘bird’; Paul 1998: 6, 11).

As is the case in Kurdish – and perhaps even more so – emphatics are part of the periphery of Zazaki phonology, appearing in a small set of words and exercising a limited functional load.

4.3 Dezfuli-Shushtari

Dezfuli and Shushtari are closely related Southwestern Iranian varieties spoken by the inhabitants of Dezful and Shushtar, cities in northern Khuzestan Province, Iran. These cities are situated in a linguistically heterogeneous area where Arabic, Khuzi Persian, Southern Kurdish and two Luri languages (Northern Luri and Bakhtiari) are spoken alongside one another. Of these contact languages, the Luri languages are the closest relatives of Dezfuli-Shushtari (Windfuhr 2009: 13).

The consonant inventory of emphatics in Dezfuli-Shushtari differs from that found in Kurdish and Zazaki in that it lacks co-articulated emphatics (ṣ, ṭ, etc.). It does, however, contain both of the pharyngeal consonants found in Arabic, ḥ and ʕ. Most of the words in which pharyngeals occur here are identifiably borrowed from Arabic (Dezfuli-Shushtari data from MacKinnon 1974: 23–34):

MacKinnon mentions that pharyngeals “occasionally also occur in words not of Arabic origin”, and cites the item ʕas ‘bone’ (1974: 12, 1995; Nirumand [1970]: 181 gives the form ʕass). ^[6]Nirumand’s (1970) short lexicon of Shushtari reveals their distribution in a number of other words of non-Arabic origin. Some of these words appear to be lexical innovations, and a small number can be traced to introduction of ʕ into inherited Iranian vocabulary. Both types of words are shown in the following list (Shushtari data from Nirumand [Circa 1970]: 151–153, 181–183):

4.4 Keshmi

Keshmi (or Qeshmi) is a Southwestern Iranian variety spoken on Qeshm Island near Bandar Abbas, Iran, at the western opening of the Strait of Hormuz. It shares many traits with Bandari (coastal) Persian, but substantive documentation is lacking. Here, as in Dezfuli-Shushtari, the pharyngeal consonants ḥ and ʕ are found, and co-articulated emphatics are absent from the consonant inventory. However, even the pharyngeal consonants are now being lost in the phonology of younger speakers and most often replaced by their glottal counterparts h and ʔ (or, in the case of ʕ at a word boundary, disappearing completely). In a 240-item wordlist collected from an older speaker of the Dargahān dialect of Keshmi in 2009, most words of Arabic origin retained the two pharyngeals (Anonby 2016). This is evident in the following list, in which Keshmi words are shown along with their Arabic source words and Persian cognates:

In two cases, however, the pharyngeal is absent in words of Arabic origin even among older speakers:

In addition, the pharyngeal ʕ appears in one Keshmi item for which, as for the word ʕas(s) ‘bone’ in Dezfuli-Shushtari (‎4.3), the determination of origin is complex. At first glance, the item ʕawr ‘cloud’ appears to be descended from Avestan awra or Middle Persian abr (note that the b ~ w correspondence has been unstable in Persian, and a b>w sound change is widespread in related Southwestern Iranian languages); if this explanation were correct, it would provide an example of a pharyngeal arising spontaneously in Keshmi. However, a similar word for ‘cloud’ ʕayba is found in Classical Syriac Aramaic (Nöldeke 1896 in Hoberman 1985: 225), and the Arabic items ʕabra(t) ‘teardrop’ and ʕabar ‘to cross, to pass over’ cannot be discounted as possible sources.

In sum, pharyngeal consonants are confined to the outer periphery of the phonological system in Keshmi, as is the case in Dezfuli-Shushtari.

4.5 Pashto

Of all the Iranian languages where pharyngeals have been documented, their distribution is most restricted in the East Iranian language Pashto. Parallel to the situation in Dezfuli-Shushtari and Keshmi, the pharyngeal consonants ḥ and ʕ are retained in words of Arabic origin. Here, however, these consonants have an acrolectal status: they are confined to the formal speech of educated speakers of the language (Robson and Tegey 2009: 725). Examples of Pashto words in which these fricatives appear are ḥabīb ‘dear (n.)’ (= Ar.) and ʕaqəl ‘wisdom’ (< Ar. ʕaql) (724). As Agnes Korn (pers. comm. 2012) has noted, this situation likely mirrors learned pronunciation of Arabic loanwords within many other languages in the broader Islamic sphere.

4.6 Emphatic consonants in Kumzari

In the five Iranian languages described so far, it is evident that while emphasis has been incorporated in various ways and to various degrees as a result of contact with Arabic, it remains in each case a peripheral aspect of the phonology. In contrast, emphasis has been profoundly phonologized in Kumzari, both in relation to phonological inventory and to the behaviour of emphasis in the language (Section ‎5).

The Kumzari emphatic series is comprised of the following six consonants:

In addition, three other consonants tend to trigger emphasis in other consonants diachronically, although they are not themselves contrastively emphatic:

The identification of emphatic consonants in Kumzari is a delicate procedure because there is no single defining structural trait which accounts for all of the elements in the system. There is, however, a set of interrelated phonetic and phonological characteristics that are useful in describing the series. These characteristics are as follows:

1. With one exception, Kumzari emphatics have contrastive non-emphatic counterparts. Contrast between members of the two groups of consonants is evident from the following pairs of words:

(Sound files with contrasting emphatic/plain consonants pairs, reproduced from Anonby 2011a, are found in the Appendix and audible in the online version of this article.)

To be thorough, the emphatic / non-emphatic contrast is weak (but still defensible) for two of these pairs. In the case of the lateral approximants ḷ and l, the emphatic member ḷ is poorly attested in the Kumzari lexicon, being found in only a small number of stems, including waḷa ‘or’, and ʔaḷḷa ‘God’ (along with some but not all of the compounds containing this stem; see Section ‎5.1 below). Concerning ḥ and h, the inverse is true: ḥ is extremely common, but its non-emphatic counterpart h is peripheral in the lexicon: it is found in a small set of words, most or all of which are interjections and/or borrowings from Arabic:

In the case of ẓ, however, a plain counterpart is completely absent from the consonant inventory as a result of a surprising sound change (cf. Table 1 in Section 2; see also Section ‎5.4). Earlier accounts of the alveolar fricative in Kumzari make reference to a plain z (Jayakar 1902; Thomas 1930), but Anonby (2011a: 376) describes it as emphatic, and this assertion will be developed throughout the remainder of the article.

1. Uvularization is the main articulatory basis for emphasis in Kumzari. The alveolar emphatics ṭ ḍ ṣ ẓ ḷ exhibit strong, simultaneous posterior secondary articulation, with uvularization dominating but bounded by a unified stricture all the way from the pharynx up to the velum. Phonetically, this complex articulation can be symbolized as [tˤ dˤ sˤ zˤ] and [lˤ]. ^[7] The remaining member of the Kumzari emphatic series, however, is a pharyngeal consonant ḥ [ħ]. Although ḥ is not uvularized, its behaviour suggests that it should be classed as an emphatic (Sections ‎5.3.1–‎5.3.2). The status of the retroflex alveolar approximant r is problematic for the opposite reason: while it does not pattern as an emphatic consonant by inducing emphasis on other consonants (Section ‎5.3), it is noticeably uvularized ([ɻˤ]) by most speakers in its most frequently attested positions, especially in word-initial and word-final positions:

This compares with the non-uvularized realization of r in other positions (Anonby 2011a: 377):

Still, as these data illustrate, and in opposition to the contrastively uvularized emphatic series outlined in the first point, uvularization of r is entirely allophonic.

The uvularized alveolar emphatics, uvular consonants x q ġ, the pharyngeal ḥ, and the uvularized allophone of r all cause preceding as well following non-back vowels to be retracted (ā [aː] → [ɑː], a [ɐ] → [ʌ]). In the case of non-low vowels, they cause lowering in the transition between the vowel and consonant (ī [iː] → [i͡ə] before a consonant, [ə͡i] after a consonant; ē [eː] → [e͡ə] before a consonant, [ə͡e] after a consonant).

1. As stated above, w, x and q are three consonants in the Kumzari inventory which, although they are not themselves contrastively emphatic, have diachronically triggered emphasis in the non-emphatic counterparts of the other members of the series (‎5.3). This behaviour supports their designation, noted for other languages with x and q, as opaque emphatics (‎3.4). There are no cases where the voiced uvular fricative ġ has been observed to trigger emphasis on neighbouring consonants, but as ġ is not abundant in the lexicon, this could reflect an accidental gap.

   Notably, the approximant w is labial-uvular, i. e., produced by a simultaneous bilabial and uvular double articulation, rather than labial-velar. In addition to inducing emphasis, the approximant w – along with x, q and ġ (but not velar consonants k and g) – causes retraction of non-back vowels (especially ā; cf. recording 6 in the Appendix) and centralization of non-low front vowels.

A summary of the emphasis-related properties of consonants in Kumzari, as outlined in the preceding discussion, are summarized in Table 2:

This complex situation underlines the reality that place and/or manner of articulation are not in themselves sufficient as predictors of which consonants participate in the language’s system of emphasis.

It is true that, like Kumzari, some Kurdish (and possibly Zazaki) varieties exhibit a highly differentiated inventory of emphatic consonants (Sections ‎4.1 and ‎4.2). However, similarities in the degree of phonologization end here.

First of all, the sheer number of emphatics in the Kumzari lexicon (Anonby and van der Wal Anonby in prep.) is remarkable: in a database of 4400 items, there are about 1200 instances of emphatic consonants distributed among almost a quarter (1086) of the words, totalling 5.3% of all phoneme occurrences in the lexicon. While the measures are not identical, this figure compares favourably to the proportion of emphatic consonants in Arabic, where in a random text sample the co-articulated emphatics ṭ ṣ ḍ đ̣ together show a frequency of only 2.3% of individual phoneme occurrences (al-Xūli 1984 in Bakallah 2009: 423). In reality, such a comparison suggests that the proportion of emphatic consonants may be equivalent or higher in Kumzari than in Arabic.

Second, whereas the overwhelming majority of emphatics in other Iranian languages are limited to Arabic loanwords, there are hundreds of non-borrowed Kumzari words – including both identifiably inherited items as well as apparent Kumzari-specific innovations – in which emphatic consonants have arisen (Section ‎5.2).

Third, while emphasis shows little inclination toward diffusion in the lexicon and phonology of the related languages examined above, there are numerous examples of diachronic and synchronic spread of emphasis in Kumzari, whether from emphatic consonants themselves or from other emphasis-inducing consonants (Sections ‎5.3–‎5.5). In fact, it is this firmly established phonological tendency that has led to the situation described in the previous two points, in which emphasis is not only well-attested in borrowed words, but has permeated the lexicon as a whole.

Now that the stage has been set with a discussion of the typology of emphasis (Section ‎3) and a review of its status in Iranian languages in contact with Arabic (Section ‎4), the following section (Section ‎5) will explore the question of how emphasis has arisen and, over time, become so profoundly phonologized in Kumzari. In short: what are the reasons for the contrasting outcomes of Kumzari versus related languages in similar situations, where emphasis has remained on the phonological periphery?

5.1 Retention of emphatic consonants in words adopted from Arabic

Like the Arabic system on which it is based, the Kumzari emphatic series is robust. Arabic emphatics ṭ ḍ đ̣ ṣ ḥ and ḷ are all consistently maintained as emphatics in Kumzari words borrowed from Arabic. The glottal stop is, however, found in place of the pharyngeal consonant ʕ (ʕayn), as in the neighbouring Shihhi dialects of Arabic with which Kumzari is in constant contact. The following paragraphs illustrate the ways in which these consonants have been retained and, in some cases, altered. As mentioned in the previous paragraph, Arabic forms are represented by Modern Standard Arabic unless stated otherwise.

ṭ, ṣ, ḥ and ḷ

In the case of ṭ ṣ ḥ and ḷ, the primary place of articulation of the Arabic source consonant has been preserved.

There are only a small number of Kumzari stems with ḷ. Examples in the data include waḷa ‘or’ (see Section ‎5.3.2); ṣāḷaṭ ‘caught hiding’ (cf. Ar. sallaṭ); and ʔaḷḷa ‘God’, where ḷ is retained from Arabic (ʔaḷḷāh). It appears in many expressions that use this root, however:

In two common expressions containing ʔaḷḷa, l is not emphatic.

The Kumzari patterning of emphasis with the root for ‘God’ reflects the Arabic source, where l is non-emphatic in this word as a result of an allophonic alternation in the environment of a high vowel (Holes 2004: 95). ^[8]

ḍ and đ̣

While the Arabic emphatic ḍ is reflected as ṭ in what appear to be older borrowings into Kumzari, it is unmodified in all recent borrowings.

The borrowing of Arabic đ̣ has resulted in the same pattern in Kumzari: ṭ in what appear to be older borrowings, and ḍ in all recent borrowings.

This parallel patterning reflects the fact that in Arabic dialects other than Classical Arabic (CA) and Modern Standard Arabic (MSA), there is no distinction between ḍ and đ̣; in Old Arabic, there was probably a single consonant corresponding to CA/MSA ḍ/đ̣, and in contemporary Arabic dialects, both phonemes are typically represented by a single consonant (usually ḍ, đ̣ or ẓ; see Holes 2004: 70–73, and ‎5.4.2).

ʕ (ʕayn)

The pharyngeal consonant ʕ is the only member of the Arabic emphatic series which is not retained in words borrowed into Kumzari. Instead, it is consistently rendered as a glottal stop ʔ.

An important consideration which helps account for the retention of all Arabic emphatics in Kumzari except ʕ is that Shihhi Arabic, a cluster of Arabic dialects which surround Kumzari (see beginning of Section ‎5), also lacks ʕ in its consonant inventory and has similarly replaced it with the glottal stop (Bernabela 2011: 26).

Rendering of Arabic consonants and time depth of lexification

Although the present article is restricted in scope to the patterning of emphatic consonants in Kumzari, even the limited data here provide a window into the time depth of the language’s initial lexification (large-scale borrowing of vocabulary), from Arabic and the resulting adoption of emphatics in the phonology.

Almost all Iranian languages were lexified by Classical Arabic (as a written language) and Mesopotamian Arabic (as a spoken language) via Persian in the period following the Islamic conquests of south-western Asia beginning in the seventh Century a.d. (Sadeghi 1986; Windfuhr 2009: 419), but Kumzari was not affected by this event. Rather, it has been separately lexified by Peninsular Arabic at some point before this – in other words, more than 1300 years ago. There are almost no New Persian loans into Kumzari, and whereas Persian (and, through it, almost all other Iranian languages) has dispensed with emphatic consonants in Arabic loanwords, Kumzari has consistently maintained them. There are also consistent differences in the primary place of articulation of the emphatic pair ḍ/đ̣: whereas the Iranian reflex z mirrors the fricative articulation ẓ in many northern Mesopotamian Arabic dialects (Jastrow 1980: 142; Sadeghi 1986: 29–30), the Kumzari reflex is consistently a stop (ṭ or ḍ, depending on the date of the borrowing; see in this section above), in line with the “City” type dialects of the Arabian Peninsula (Section ‎5.4.2; Holes 2004: 71). This correspondence is shown in Table 3.

Additional patterns of Kumzari vs. Iranian borrowing and their implications for the linguistic history of Kumzari are developed in greater detail in Anonby (in prep. a).

5.2 Propagation of emphatic consonants through individual innovations

Once emphasis has entered a language through retention of emphatic consonants in loanwords, it may spread into new contexts (Matras 2009: 222–224). In Kumzari, emphasis sometimes appears in individual words, and there is no discernible pattern. In other cases, spread is achieved through systematic changes. Both types of innovations highlight the degree to which emphatic consonants have been incorporated into the core of the phonological system (see analogous examples of loan phonology integration in Aikhenvald 2006: 21, 23; Windfuhr 1990: 543–544). While the following sections will address systematic patterns of spread, the present section will first account for spontaneous appearances of emphasis through lexical innovation (Section ‎5.2.1) and, in a handful of cases, on potentially emphatic consonants in existing vocabulary (Section 5.2.2).

5.3 Diffusion of emphasis through analogical sound change

An additional outcome made possible by the initial phonologization of emphasis in Kumzari is the recurring diffusion of emphasis onto potentially emphatic consonants (Section ‎3.4) within a word. This diachronic process operates by means of “lexical diffusion” of analogical sound change (Chen and Wang 1975; Kiparsky 1995): there are specific phonological contexts within which emphasis is likely to spread, but it is not automatic; rather, it diffuses on a word-by-word basis. Changes have taken place in a significant and apparently growing number of words in the language, and apply equally to inherited vocabulary and Arabic loanwords.

The first half of this section explores the phonological conditions and mechanisms for the analogical diffusion of emphasis (Section ‎5.3.1). In the second half, examples of diffusion are catalogued for each of the specific sound changes brought about by this process (Section ‎5.3.2).

5.4 The sound change z > ẓ

The phonologization of emphasis has reached its acme in Kumzari with the implementation of an across-the-board sound change – applying without exception – in which z has been recast as an emphatic ẓ; non-emphatic z has now disappeared from the consonant inventory. The z>ẓ sound change has applied throughout the Kumzari lexicon, affecting inherited vocabulary as well as words borrowed from Arabic. Examples of inherited vocabulary in which this innovation has taken place in Kumzari are as follows:

Examples of Kumzari words originally borrowed from Arabic, and in which z has similarly undergone this sound change, are as follows:

The displacement of z by its emphatic counterpart ẓ equally applies to Kumzari words recently borrowed from English via Gulf Arabic, such as flēẓar ‘freezer’ and talfaẓūn ‘television’.

The existence of an emphatic ẓ with no plain counterpart is an example of the extraordinary situation in which a marked phoneme is found in the language without an unmarked counterpart. Specifically, this goes against the universal tendency that segments with secondary articulations are only found when there is also a plain counterpart (Gamkrelidze 1978: 11–13).

Since historical z was a common phoneme in both inherited and borrowed vocabulary, the effects are far-reaching in present-day Kumzari: the resulting emphatic ẓ appears in 327 items (more than 7% of words) in a lexicon of just under 4400 entries. This compares, for example, with about the same number of items (335 words) containing f, and 217 items (5%) containing j [d͡ʒ], when both of these other sounds are manifestly better represented cross-linguistically than ẓ. This pattern runs counter to the cross-linguistic tendency for rare phonemes, and in particular co-articulated emphatics, to be poorly represented in the lexicon (Delattre 1971: 131; Gamkrelidze 1978: 13).

This typological anomaly is addressed here by means of three questions. First, in terms of the Kumzari phonological system, would it be more appropriate to interpret [zˤ] as a plain consonant? (Section ‎5.4.1) Second, if it really is an underlyingly emphatic consonant with no plain counterpart, can it be attributed to borrowing from Arabic? (Section ‎5.4.2) And third: if not, how has such a situation arisen? (Section ‎5.4.3). The discussion closes with reflections on the ultimate motivation of this sound change (Section ‎5.4.4).

5.5 Synchronic alternation

In large part, the spread of emphasis in Kumzari is a diachronic phenomenon (Sections ‎5.2–‎5.4). But synchronically, there are two contexts in which emphasis alternates systematically across morpheme boundaries. ^[11]

First, when the “count” morpheme -ta is added to the bare form of numbers that end in an emphatic consonant, the morpheme-initial t becomes emphatic ṭ. There are only a handful of contexts where this process can take place (i. e., number stems ending in emphatic consonants), and other morphophonological alternations take place at this boundary in many cases, but a comparison with stems ending in plain consonants makes the scope of its application clear:

cf. other consonant-final numbers:

In a second context, but only with some speakers – perhaps because the co-articulation of emphasis is in the process of phonologization here – when noun stems ending in an alveolar emphatic consonant are inflected with the comparative suffix -tar, the initial t of the suffix similarly becomes emphatic ṭ.

As may be expected, the t of the suffix does not become emphatic when the stem does not end with an emphatic consonant.

Surprisingly, however, this process does not take place with the pharyngeal consonant ḥ, or with the emphasis-inducing consonants q, w and x.

To frame this second process in the terms of the Arabic grammarians introduced above, the synchronic spread of emphasis in Kumzari is limited to consonants which fall under the narrow definition of emphasis as ʔiṭbāq (Section ‎3.1), that is, alveolar consonants with a secondary posterior articulation. This stands in contrast to the diachronic diffusion of emphasis in Kumzari, which operates according to broader conceptualization of emphasis that includes other posterior consonants: the pharyngeal ḥ and the uvulars w, x and q, that (Sections ‎3.4 and ‎5.3.1).