1 Consonants

STKF claim that laryngeal oppositions within obstruents in Halh are between values of [±spread glottis] rather than [±voice].Footnote ² They endorse Ramstedt's (Reference Ramstedt1903) observations that the fortis series is postaspirated in word-initial position and preaspirated elsewhere. The lenis series are ‘ordinary voiceless unaspirated stops, similar to those found in Russian or French’ (p. 12). The writing system for Mongolian used in the Republic of Mongolia is the Cyrillic alphabet, and the obstruents are written with the voiceless and voiced symbols in Cyrillic, e.g. <T> denotes [tʰ] and <д> denotes [t]. Like Icelandic (Thráinsson Reference Thráinsson1978), vowels preceding preaspirated stops are partially devoiced (p. 14).

Aspirated consonants do not undergo passive voicing when vowel initial suffixes are added, e.g. [xɔtʰ] ‘town’, [xɔtʰɔs] ‘town-abl’. (In this way they are like the fortis stops of Turkish, in the analyses of Avery Reference Avery1996 and Petrova et al. Reference Petrova, Plapp, Ringen and Szentgyörgyi2006). STKF claim that /s/ is [+spread glottis] (p. 18; cf. Iverson & Salmons Reference Iverson and Salmons1995, Vaux Reference Vaux1998). The fricatives [s ʃ] are postaspirated (as is clearly shown in spectrograms on p. 19).

Velar [ŋ x] vs. uvular [N χ] are in complementary distribution, according to the [±Advanced Tongue Root] ([±ATR]) harmony value of the word as a whole, e.g. [ʧooŋ] ‘few’ vs. [ɔN] ‘year’ (pp. 28–29). Velar /g/, however, may occur in morpheme-final position in pharyngeal (e.g. [−ATR] harmonic) words, e.g. [pag-as] ‘team-abl’ (vs. [paGas] ‘small-abl’). Halh has the lateral fricative /ɮ/, but no plain /l/, and voiced /g gʲ g/ but no /k kʲ q/. The glide /w/ developed from Old Mongolian *p postvocalically, hence w is mostly not found initially (though in loanwords e.g. waar ‘tile’) from Chinese) and p is mostly only found initially and after [m w ɮ] (p. 29). They can contrast underlyingly /aɮp/ ‘service’ vs. /aɮ-w/ ‘to kill-past’, but on the surface the latter will undergo epenthesis, becoming [aɮəw]. A series of palatalised consonants (/pʲʰ tʲʰ pʲ tʲ gʲ xʲ mʲ nʲ lʲ rʲ wʲ/) derive historically from a following *i (p. 28), to be discussed further below.

2 Vowels

Halh contrasts long and short vowels (though there is no short /e/, and short /o/ may be closer to [ɵ]; p. 4). According to STKF, long vowels are found only in initial syllables (p. 22). Although STKF do not commit themselves as to where main word stress is (‘our conclusion is that word stress is not phonologically relevant in Mongolian’ (p. 96)), one might infer that since the long–short vowel contrast is only in the initial syllable, this is where stress is located.Footnote ³

In non-initial syllables the distinction is not between long and short, but rather between full and reduced vowels. While full vowels in non-initial syllables derive historically from long vowels, in Modern Halh they are equal in duration to initial short vowels (p. 3). Thus, even though they are written as doubled vowels in Cyrillic (e.g. MuHuŭ ‘my’), they are represented as short by STKF: [mini] (p. 25). Reduced vowels in non-initial syllables are centralised variants of the vowel in the preceding syllable (p. 6), unless preceded by an alveopalatal sibilant or a palatalised consonant, in which case they are [i]-like (p. 23). STKF claim that these reduced vowels are inserted epenthetically (ch. 6), and thus Cyrillic [aʧiɮ] ‘work’ is underlyingly /aʧɮ/ (p. 25).Footnote ⁴

1. (1)

   

The most important division in the vowel system is between [+ATR] [u e o], [−ATR] [ʊ ɔ a] and neutral [i]. (STKF use privative [pharyngeal], but I will use [±ATR] here.) The Halh [+ATR] vowels have a higher F1 and lower F2 than their [−ATR] counterparts, and articulatorily results from a wider pharynx (based on x-ray evidence in Cenggeltei & Sinedke Reference Cenggeltei and Sinedke1959), leading to an articulatory basis similar to what Lindau (Reference Lindau1979) proposes for such contrasts. This division between the vowel sets is relevant for vowel harmony, and also affects the distribution of velar vs. uvular consonants, as well as palatalised variants of many of the consonants.

The basics of [±ATR] harmony are shown below, with the causative suffix.

1. (2)

   

Rounding harmony occurs only among the [−high] vowels. It is shown here for the past tense suffix, which is underlyingly [−high].

1. (3)

   

Interestingly, high [+round] vowels block harmony from occurring, even when a [+round, −high] vowel is further to their left.Footnote ⁵

1. (4)

   

However, the vowel /i/ is transparent to both ATR and rounding harmony, as can be seen in the following examples.

1. (5)

   

In addition, with the diphthongs /ai ɔi ʊi ui/, only the nucleus is relevant for vowel harmony, confirming this as another case of /i/-transparency, since this /i/ is demonstrably a vowel and not a glide (see below). However, STKF claim on p. 9 that the realisation of /i/ is different after [+ATR] vs. [−ATR] words in their F1 and F2 values.Footnote ⁶ Stems with only /i/, or with /i/ in the initial syllable, are [+ATR] (p. 192).

Velar [g] and uvular [G] are epenthesised before vowel-initial suffixes (p. 55), depending on the harmony value of the word. STKF claim that dorsal is the unmarked place for consonantal epenthesis (which raises interesting issues; cf. Blevins Reference Blevins and Good2008), perhaps since it is already part of the feature specification of all vowels (cf. Howe Reference Howe2004).

1. (6)

   

Interestingly, however, when the vowel immediately following the dorsal is /i/, the consonant has a velar realisation, even in [−ATR] harmonic words (p. 55): cf. [sana-gin] in (6). This suggests that the [g/G] alternation is determined by the following vowel, and that [i] is in fact [+ATR], even though it is transparent for harmony.Footnote ⁷

3 Syllabification

Halh disallows complex onsets, but allows superheavy syllables, such as [nuursʧʰ] ‘coalman’, [ʊʊrstʰ.ɮa] ‘to be angry-term-refl’ (p. 64). The consonants /ŋ ɮ w ɾ/ are not allowed word-initially (p. 26), and /ŋ/ is not allowed syllable-initially, changing to [n] in case of resyllabification (p. 63).

Halh has a bimoraic minimal word requirement (p. 78), satisfied either by coda consonants or long vowels. It allows two-consonant codas that obey a sonority decrease, and has place assimilation of nasals in codas, except for [m], which can occur before coronal stops. It allows three-consonant codas where the first element is either a sonorant or /g/ (but not /G/), the second is a fricative and the third a coronal stop, such as [ʧimstʰ] ‘having fruit’, [ʃarxʧʰ] ‘coroner’, [tsigsʧʰ] ‘warbler’ (p. 68).

Epenthetic vowels are inserted ‘as far to the left as possible’ (p. 69), e.g. [Gʊ.təmʧ] ‘street’, not *[Gʊt.məʧ], from /Gʊtmʧ/, [Gʊ.rənts] ‘emery’, not *[Gʊr.nəts], from /Gʊrnts/.Footnote ⁸ Other examples where the last two consonants cannot form a coda include [xit.məŋ] ‘pear’, [ʧʰotʰ.gər] ‘devil’ and [gur.wəɮ] ‘lizard’. Schwa–zero alternations provide additional evidence for an epenthesis analysis. The nine successively morphologically derived words in (7) show this (p. 71); fleeting schwas are underlined.

1. (7)

   

In the examples above, since /gʧʰ/ can never form a complex onset, the schwa before it stays, but other schwas are deleted as soon as coda consonants become resyllabified as onsets.

The apparent complete predictability of schwa placement is challenged by minimal pairs such as [tso.wəɮ.ɮo] ‘advise-dir past’ vs. [tsow.ɮə.ɮo] ‘his advice’. STKF propose that their morphological structure is responsible for the difference in syllabification (p. 73), with a cyclic constraint prohibiting schwa deletion from already-formed stems (cf. Bobaljik Reference Bobaljik and Samiian1997 for related discussion). Thus, [tso.wəɮ] is syllabified as a verb, with the direct past added, forming [tso.wəɮ.ɮo], with no loss of the schwa from the verbal stem. By contrast, [tso.wəɮ] becomes a noun due to the nominalising suffix [-ɮ], causing schwa deletion and insertion at the end: [tsow.ɮəɮ]. Subsequent addition of the possessive cannot induce deletion of the schwa from the nominal stem: [tsow.ɮə.ɮo].

Halh differentiates /i/ and the palatal glide /j/ in underlying representations (cf. Levi Reference Levi2004). These are distinguished in orthography as glide-final vs. diphthongal rhymes (cf. ɑя [aj] vs. ɑŭ [ai] below), and according to STKF, offglide sequences have steeper formant movements than i-diphthongs (p. 11). For example, the glide-final words in (8a) differ from the ones in (8b) in that the latter trigger hiatus-breaking [G]-epenthesis with the instrumental suffix -Er (p. 77).

1. (8)

   

In addition, in careful pronunciations, the offglide sequences trigger schwa epenthesis when a vowel follows, e.g. [pair] ‘place’ vs. [pajər] ‘happiness’ (p. 78).

4 Historical phonology

Old Mongolian had the vowel system in (9), with [±back] harmony between /y~u/, /ø~o/ and /e~a/, with neutral /i/ (p. 114). Old Mongolian had no rounding harmony.

1. (9)

   

STKF posit that the first stage of its vowel shift was pharyngealisation of the back vowels, whereby /u o/ became /ʊ ɔ/.Footnote ⁹ This created a ‘vacuum’ for the vowels /y ø/ to undergo backing, becoming /u o/, with the two processes together composing a kind of pull chain. Modern Halh /u o/ thus correspond to Old Mongolian (OM) *y ø and /ʊ ɔ/ to *u o. Kalmuck did not undergo any of these changes, and Shira Yugur did not have backing of *ø. Dagur underwent a further step than Halh, merging /o/ with /u/ and /ʊ/ with /ɔ/ (p. 180). These four daughter languages are compared below.

1. (10)

   

As a result of these developments, Kalmuck retains [±back] harmony, while the rest now have [±ATR] harmony (see Svantesson Reference Svantesson1985, which corrects the majority of earlier theoretical literature, which assumes that Halh harmony is [±back]).

OM /i/ in initial syllables underwent an interesting process of assimilation to the following vowel in words with following [+back] (corresponding to Halh [−ATR]) vowels, leaving palatalisation of the preceding consonant, e.g. OM *kʰilpar>Halh [xʲalpər] ‘easy’; OM *siŋkʰor>Halh [ʃɔŋxər] ‘falcon’ (p. 195).Footnote ¹⁰ Dagur underwent a similar process with initial /u/, yielding labialised consonants: OM *tumta > Dagur [tʷant] ‘middle’; OM *kurpan > Dagur [kʷarpə] ‘three’ (p. 197). Non-initial /i/ in [+back] harmonic words was deleted entirely, leaving a trace of palatalisation: OM *tʰapin ‘fifty’ > Halh [tʰawʲ], OM *alima ‘fruit’ > Halh [aɮjəm] (p. 209). As a consequence of the general loss of non-initial vowels, STKF mention that while Old Mongolian had no complex codas, Halh has 157 different complex codas, and while two-thirds of Old Mongolian words were bisyllabic, two-thirds of Halh words are monosyllabic.

In many dialects (though not Halh), Old Mongolian aspirated stops underwent a Grassman's Law-like process of deaspiration when an aspirated stop or /s/ occurred in the following syllable. In Chahar, OM *tʰatʰa ‘to pull’ > Chahar [tatʰ] (cf. Halh [tʰatʰ]); OM *tʰosun ‘fat’ > Chahar [tɔs] (cf. Halh [tʰɔs]) (p. 206). Perhaps most interestingly, Monguor displays an aspiration ‘flip-flop’, whereby unaspirated–aspirated sequences become aspirated–unaspirated: OM *totʰara ‘inside’ > Monguor [tʰutor], OM *patʰu ‘firm’ > [pʰati]; OM *pyse ‘belt’ > Monguor [pʰusee] (pp. 207–208). As STKF note, this flip-flop might result if the partial devoicing effect that preaspirated stops have on preceding vowels (mentioned above in §1) could be reinterpreted as postaspiration of the initial consonant (cf. Ohala Reference Ohala, Masek, Hendrick and Miller1981). Again, the fact that /s/ patterns with the aspirated stops in all of these processes is further evidence of its [+spread glottis] character.

5 Loanword phonology

As a result of the consonantal inventory and phonotactics of Halh, loanwords from Russian can undergo a number of processes of phonological restructuring. Since there is no initial /r/, loanwords such as [ˈradʲiɐ] ‘radio’ are usually pronounced with an initial vowel ([araʧəw]; p. 30). Russian [f] becomes /pʰ/, e.g. [fɪˈgʲurɐ] ‘figure’ > [pʰigur] (p. 31). [k] is sometimes retained, but may also be changed to /x/ ([ˈxasɐ] ‘cash-desk’ > [xaas]) or to /g~G/ ([kəlbɐˈsa] ‘sausage’ > [Gaɮəwsa]).

Stressed vowels in Russian become long vowels in Mongolian (e.g. long in initial syllables and full/phonemic in non-initial syllables; p. 32). They also determine the vowel harmony class of the word. Russian o [ɔ] thus decides the harmony class of words such as [vɐˈgɔn] ‘coach’ > [pɔGɔŋ] (unless palatalised consonants determine that the harmony class is [−ATR], as in [mʲɪˈnʊtɐ] ‘minute’ > [mʲanʊtʰ]). Initial clusters are broken up by a copy of the stressed vowel (e.g. [ˈbraʃkɐ] ‘mash’ > [paraʃək], [ˈplʲitkɐ] ‘stove’ > [pʰiɮitʰəg]), but if it begins with /sC/, an initial /i/ may be inserted (e.g. [ˈspʲirt] ‘alcohol’ > [ispʰirtʰ], [ˈʃkaf] ‘cupboard’ > [iʃkaw]) (cf. Fleischhacker Reference Fleischhacker2001 for discussion of similar patterns).

Final unstressed vowels are deleted (e.g. [ˈdramɐ] ‘drama’ > [taram]), and unstressed vowels are deleted when not required for syllable structure (e.g. [ɐftɐˈmat] ‘slot-machine’ > [awtʰmatʰ]). Perhaps most interesting is loss of [i] after palatalised consonants, e.g. [lʲimɐˈnat] ‘lemonade’ > [ɮʲamnat], [mʲɪˈnʊtɐ] > [mʲanʊtʰ], consistent with the phonotactics of Halh, in which Cʲ is not found before [i].

6 Reduplication

Halh has fixed-segment coda w (p. 58) for intensive adjective reduplication.

1. (11)

   

Vowel length and the second half of diphthongs are ignored in the reduplicant,Footnote ¹¹ leading STKF to adopt a CV-slot analysis, along the lines of Marantz (Reference Marantz1982). Note that the fixed segment is a p in the Turkish equivalent (Kelepir Reference Kelepir, Göksel and Kerslake2000), as well as in Eastern Mongolian dialects (p. 59), and recall that Halh postvocalic *p has gone to /w/.

Nouns can form echo reduplication, with the associative plural semantics (‘X and such things’, ‘X and people like him/her’, with a slightly pejorative flavour). This is formed by an m- prefix that appears in the onset of the reduplicant (12a), unless the base begins with m, in which case it is [ts] (12b). This process cannot be treated as complete overwriting of the onset in the reduplicant, because of the interesting fact that palatalisation is transferred from the corresponding consonant in the base (p. 60) when m- is chosen, resulting in [mʲ] (12c). However, as /ts/ has no palatalised counterpart, no transfer occurs when [mʲ] is the base (12d).

1. (12)

   

The statement of palatalisation preservation requires a representation in which the secondary articulation of the base is preserved even under replacement of the primary features of the consonant (as STKF note on page 61, this supports a representation where palatalisation is on a separate tier), and is deleted if incompatible with the primary articulation.Footnote ¹²

7 Concluding remarks

A large portion of the book consists of comparative tables of words in Old Mongolian and the eleven modern Mongolic languages, as well as tables showing vowel and consonant developments in schematic format. In my opinion, an accompanying CD or website would make these much easier to traverse and search for particular patterns.

The authors refrain from developing theoretical models of many of the phenomena (in contrast to most other volumes in this series), but they do organise the presentation of the phenomena in a way that makes them accessible for interested researchers of all stripes. In addition to its own important empirical contributions, given the paucity of literature on the phonology of Mongolian in English, this book is a useful synthesis of much existing literature (it contains twelve detailed pages of annotated notes on sources and literature written in Mongolian, Russian, Chinese and other languages). The sketches of the non-Halh Mongolic languages are limited but informative. The authors do not attempt a genetic subclassification of the Mongolic languages, doubting the appropriateness of a family tree model for this, in light of the complicated history of language contacts (p. 217).

The book as a whole is organised in a way that is challenging for a linear reading (e.g. there is a presentation of Kalmuck orthography in Chapter 4 before we know anything about Kalmuck; there is no discussion of stress until Chapter 7; presentation of loanword phonology – which could have been a chapter in its own right – comes before the epenthesis rules are introduced). It is worth the effort, however: many of the phenomena mentioned above have not yet been given a full treatment in contemporary phonological theories of vowel harmony, syllabification, reduplication or loanword phonology, making this book a ripe source for interested researchers.