< 1 >

The data in this paper was obtained during a discovery workshop in April-May 2012 in Khartoum, Sudan, attended by mother-tongue writers from the following four Talodi languages: Lumun [lmd], Acheron [acz], Tocho [taz], and Dagik [dec]. The workshop participants created grammar booklets on their languages in 12 days by completing data tables for a different section of the survey each day. I then re-transcribed the data from all four languages in the International Phonetic Alphabet to enable linguistic comparison here. This documentation process was characterised by three particular methodological features which I will briefly review in turn: a language cluster approach, a parts of speech approach, and documenting languages with mother-tongue writers.

<2>

A language cluster approach, as described by Lewis and Stalder (2010), conducts work on several languages on some common basis, whether similar logistics (operational basis), a high degree of linguistic similarity (linguistic basis), ethno-linguistic affinity that stimulates co-operation in reaching common goals (sociolinguistic basis), or a combination of these. The cluster approach is well-suited to the linguistic geography of Africa, which contains vast numbers of situations where sets of immediately related languages are spoken in physically adjacent communities. The present work adopted this perspective to discover the grammar of four Talodi languages, which were documented in parallel in the same time it could have taken to document one of the languages. The expectation of high linguistic similarity between the related languages informed the planning of the discovery process, as data tables to be filled in were customised to reflect paradigm structures reported in previous sources on Talodi languages, so that these structures could be checked in the four languages present. Documenting the languages in parallel then produced situations where a consistent property was discovered in most or all of the four languages on the same day. Typically, this is a common inheritance shared by the languages, and thus reinforces the reliability of the finding.

<3>

A parts of speech approach involved collecting data in sections for different morphosyntactic word classes. This approach to grammar discovery is rooted in a typological perspective (Kröger 2012) and enables an assessment of word order type to be made (Dryer 1992). On a practical level, it takes advantage of the familiarity of parts of speech to the mother-tongue participants from their formal learning of English or other developed languages, which contributes to a discovery process that is achievable, understandable, and useable (Kröger 2012). However, it is worth emphasising that each part of speech is established by the evidence collected, not by imposing an English model. The English parts of speech are useful as a source of terminology, but word classes vary across languages (Munro 2007), and even in English they remain a matter of on-going research (Payne, Huddleston and Pullum 2010). A benefit of parts of speech research in the world’s languages is that it provides a basis for dictionary construction (Munro 2007).

<4>

The third notable feature of the documentation process is the practice of documenting languages with mother-tongue writers. This option is increasingly possible and appropriate as the number of written languages in Africa (and the world) has increased in recent decades. It has the ethical advantage of enabling research to be shared immediately with the language communities, which was achieved by the writers keyboarding their data each day to be printed as a booklet to take away at the end of the workshop. There are also linguistic advantages of working with mother-tongue writers, due to their ability to provide accurate phonemic representations that are not vulnerable to the hearing errors of a transcriber from outside the community. To ensure phonemic accuracy, languages represented in the workshop were required to have existing phonemic orthographies, and writers who were competent in using those orthographies.  [2] More specifically, writers’ data can be assumed to conform to the lexical level of phonological representation, prior to the application of post-lexical phonological changes that apply across words below speaker awareness (Mohanan 1986). This is not always the same level of transcription as used in other sources on these languages,  [3] but the lexical level is arguably the most cognitively accessible level of phonemic representation for the speakers of a language, and hence both the optimal choice for, and often spontaneously reflected in, orthographies (Mohanan 1986, Snider 2014). I thus consider it a theoretical advantage as well as a practical one that writers’ data provides a cognitively realistic representation of the language.

<5>

Working with writers also raises problems of word boundaries and tone, however. Regarding word boundaries, pioneer writers of a language sometimes justify word boundary choices by appeal to the word boundaries used between equivalent morphemes in the developed language of wider communication. Thus, mother-tongue writers’ intuitions on word boundaries in their own language seem less reliable (or are less reliably accessed) compared to their remarkable phonemic awareness. As a result, linguistic criteria for determining word boundaries come into consideration (Kutsch Lojenga 2014b). Tone presents a more serious problem: since tone is not marked in the orthographies of the languages in this survey, tone is also absent from the phonetic re-transcriptions of the writers’ data in this survey. Talodi languages appear to fit a type of African language in which tone has a low functional load in the lexicon, where the case for exhaustively representing tone in the orthography is unpersuasive (Kutsch Lojenga 2014a:63). Even if writers were to accept training to add tone marks, it could not be expected to be fully reliable (Bird 1999). However, this absence of tonal data is disadvantageous for grammatical research, because the question of functional load in the lexicon is orthogonal to the question of how tone functions in the grammar (Kutsch Lojenga 2014a:62). I therefore noted homographs in the writers’ data that may conceal grammatical tone contrasts and as such may need other forms of research. [4]

<6>

A comparative lexical study in Norton & Alaki (2015) using word lists collected at the same workshop confirmed that the four varieties in this study are separate languages, out of a total of nine Talodi languages spoken in communities in the south-eastern Nuba Mountains. That study also classifies Lumun in a separate sub-branch from Tocho, Acheron, and Dagik of the Narrow Talodi sub-branch, and the present grammatical survey provides an immediate opportunity to test this branching on additional data. [5]

<7>

When starred reconstructions are used in this paper to refer to certain morphemes shared by the languages, the reconstructions are based on the sound correspondence tables for Talodi languages given in Norton & Alaki (2015). Reconstruction to proto-Talodi requires that cognates are present in both sub-branches, and thus present in both Lumun and in at least one of Tocho, Acheron, or Dagik. However, there is a possibility that Tocho and Acheron may have acquired some structures from Lumun due to contact across their adjacent settlements at Saraf Al-Jamous, as already found in basic lexicon in Norton & Alaki (2015). Hence, the safest reconstructions here are those for cognates present in (at least) Lumun and Dagik.

<12>

Most nouns alternate in the initial consonant between singular and plural:

As previously documented in several Talodi languages by Schadeberg (1981), some nouns alternate in the first vowel as well as the initial consonant:

Some nouns have completely different roots for singular and plural:

Kinship terms use a suffix in the plural:

Some mass nouns do not distinguish singular and plural:

<13>

In the genitive construction that expresses an association between two nouns, the consonant of the head noun is repeated with a vowel *ɔ before the second noun, although /ɛ/ may be used in western (W) Acheron.

<14>

In the table above, the agreement consonant and the genitive vowel are provisionally analysed as joined to the second noun in Lumun and in Acheron, but as a separate genitive particle in Tocho and in Dagik. To decide word boundaries, one test is whether or not an initial plosive consonant on the second noun undergoes intervocalic voicing, because if voicing is a lexical property in these languages (footnote 4), intervocalic voicing would be triggered by affixation of the genitive vowel to the second noun. This test gave contrasting results in Lumun and Tocho: in Lumun, the initial consonant of the second noun is voiced, implying the genitive form /Cɔ-/ is affixed, whereas in Tocho, the initial consonant of the second noun remains voiceless, implying a separate genitive word /Cɔ/ between the two nouns. In Acheron, initial plosives are always voiced so this test does not help. However, the western dialect vowel /ɛ-/ also occurs within verbs as a relative prefix (9.4), implying that it is a prefix here also. In Dagik, voicing occurs (not in the parallel examples above, but in səŋɔ sɔ [b]urək p-abaɾɪ (singing gen person female) ‘singing of a woman’), but it applies across various words in Dagik sentences (9.4, 10, 11) so it does not establish wordhood in this language. The genitive marker is tentatively written separate from the noun here by analogy to locative markers, which are considered separate from their noun in Dagik (9.4).  [9]

<16>

Verbs may or may not take a suffix in the imperative; two imperative suffixes *-ʊ,*-ɪ both occur in Acheron and in Dagik, but in Lumun and Tocho, surveyed verbs other than ‘eat’, ‘sit’, or ‘cook’ lack suffixes. The imperative plural (that is, second person plural) is always marked by an affix, in Lumun a 2pl prefix (8.2) added to the citation stem not the imperative form, and in Tocho, Acheron, and Dagik a 2pl suffix added to the imperative form:

In all four languages, ‘go’ and ‘come’ have suppletive imperatives:

<17>

In the following verbs, the imperative has another suffix *-d̪ɛ which appears in addition to the imperative suffix if there is one, and which replaces a stem-final consonant (-t/-k) if there is one (‘give’ is more idiosyncratic, being invariant in Lumun and suppletive in Acheron):

Acheron is the only language that has either an imperative suffix or suppletive imperative for all 11 verbs surveyed, and Dagik has these for 10 of the 11 verbs. In Lumun and Tocho, few of the 11 verbs take an imperative suffix, leading to homography in several verbs between imperative and citation stem. The possibility that these contain undescribed tone changes is confirmed for Lumun imperatives in Smits (2017ː339-341).

<18>

In most of the 11 verbs, past is marked on the stem by -t in Lumun (Smits 2017ː355) or -k in Tocho and in Acheron (Alamin 2003), though no cognate suffix appears on the verbs in Dagik. In certain verbs, the suffix *-d̪ɛ is used instead, replacing a stem-final consonant (-t/-k) if there is one as first observed by Alamin (2003) in Acheron, and as also seen on imperatives in 4.1 (in this and subsequent data tables, initial p- or b- marks subject agreement with a p/b-initial subject noun or a singular pronoun):  [10]

<19>

Habitual aspect is expressed in Dagik by an auxiliary verb /a/, while the other languages have a prefix /a-/, and the auxiliary/prefix distinction is sustained in forthcoming subsections, where further inflections are added to the auxiliary in Dagik, or added to the verb containing /a-/ in the other languages. Tocho and Lumun habituals show an additional pattern of pluractional marking by internal reduplications, extensions, and suppletions relative to the citation stem shown in brackets (also in 4.6, 11.2 and Smits 2017:461ff).

<20>

In all four languages, progressive aspect is expressed by an auxiliary verb before the main verb. In three of the languages, the auxiliaries show locative extensions /-ɾɪ,-ɟɪk,-ɪk/ on the base /a/ as used for habitual aspect in Dagik.

The suffix *-d̪ɛ appears on certain verbs in the progressive in Dagik only. The additional use of the prefix /a-/ on main verbs in Tocho and Lumun, also used for habitual (4.3) or prospective (4.5), is presumably expressing a more complex inflection.

<21>

To express future time, Dagik uses an auxiliary verb with a different extension /-nta/, while the other three languages again have the prefix /a-/ giving it a broader prospective or incompletive meaning. Acheron and Tocho have additional prefixes, but in Lumun the /a-/ prefix alone gives a prospective reading. None of the additional morphemes /ɪy-/, /d̪-/, /-nta/ are cognate so they are likely to have different underlying source meanings.

<22>

Imperative verbs can be negated in all four languages by a negative auxiliary verb, which takes the imperative plural marking. In Lumun, the different stem-final vowel of the negative auxiliary in the imperative plural reflects the earlier observation that the Lumun imperative plural uses the citation stem and not the imperative stem. The Acheron data use prefixation (which recalls the use of prefixation for imperative plural in Lumun), but Acheron participants reported afterwards that they also have a negative auxiliary /d̪əma/ cognate with that of Dagik.

<23>

In declarative sentences, the strategies are more varied. Lumun uses incorporation of the negative auxiliary in reduced form, Tocho uses a postverbal negative particle, Acheron uses a negative verbal prefix, and Dagik uses the same auxiliary as with negative imperatives, with inflection expressed on a following dependent copula verb /ka/ː

<24>

In both Lumun and Acheron, the negation marker co-occurs with the incompletive prefix /a-/ (4.3-4.5) in declarative verbs. The Lumun data uses a reduced form kənn- of the negative auxiliary ɔkərənnɔ, and past is not distinguished as the past suffix appears neither after the incorporated auxiliary kənn- nor after the main verb root. This contrasts with Acheron where the past suffix does occur at the end of the verb, b-əd̪-a-uttɛ-k.

<25>

There are recurrent forms for negation across the Narrow Talodi languages. The Dagik negative auxiliary ðəma and its Acheron cognate d̪əma (reported afterwards by the participants) have the form of the Acheron negative prefix d̪- plus the Tocho negative auxiliary ma, both of which were previously reported by Stevenson (1957:39) in the Narrow Talodi language Masakin Tuwal (prefix t̪-, auxiliary maa).. The post-verbal negation particle tɪ in Tocho (becoming flapped ɾɪ after a vowel) was also reported in Masakin Tuwal (Stevenson 1957:39), and subsequently in Dagik (Vanderelst 2016ː124). Thus, three negative morphemes *t̪-,*ma,*tɪ all occur in at least three or four Narrow Talodi languages, and these languages use multiple negation strategies. In place of Narrow Talodi *ma, Lumun employs a different verbal stem ɔkə́rənnɔ ‘let, abstain’ (Smits 2017ː423ff) with negation as an extended meaning.

<26>

Imperative is especially prominent in Talodi languages in the sense of Bhat (1999) as it shows extensive morphologisation: suppletive imperatives, two imperative vowel suffixes that replace final stem vowels, a plural suffix for imperatives, and negative auxiliaries that are used only with imperatives. It is also reported elsewhere that imperatives are marked by tone changes as well (Vanderelst 2016:108-109, Smits 2017:339-341).

<27>

Past is quite prominent as well in Lumun, Tocho, and Acheron, with two suffixes marking past for different verb classes, and these also appear pervasively in published Lumun narrative texts (Smits 2007a, Stirtz 2012). At the same time, the prominence of past is limited in that Lumun does not use the past suffix on ‘give’ or on negated verbs, and the suffix *-d̪ɛ, though required by some verbs such as ‘cook’ or ‘sit’ as the only expression of past, is required in other inflectional categories as well, and Smits (2013) has analysed this suffix as a locative-applicative extension. Like the other prominent category of imperative, past is grammaticised by suffixation, in contrast to other inflections that are found before the verb root in prefixes and auxiliaries.

<29>

In all four languages, adjectives follow the noun and agree with the noun in the initial consonant in both predicative and attributive use. In predicative use, adjectives come after a copula verb that inflects like other verbs. The same copula verb root *ɡa occurs in all four languages, but is left out in the unmarked present tense in Tocho, Lumun, and Dagik:  [11]

<30>

Predicative adjectives in present tense (‘the goat is black’, ‘the goat is big’) occur after a copula verb in Acheron but are unmarked in Tocho, Lumun and Dagik. Attributive adjectives (‘black goat’, ‘big goat’) are unmarked in eastern Acheron, but in other languages are marked by relative prefixes, western Acheron -ɛ- (before a consonant), Tocho ɔN-, Lumun and Dagik ɪ-. So these attributive constructions can be interpreted as nouns modified by relative clauses containing an adjective predicate (‘goat which is black’, ‘goat which is big’):  [12]

<31>

In all four languages, the size adjective roots ‘small’ and ‘big’ take quantity meanings ‘few’ and ‘many’ in the plural except when reduplicated, as first reported in Tocho by Schadeberg (1981:132,148) (reduplicated forms are connected by ‘~’ following the Leipzig Glossing Rules, instead of the usual morpheme boundary symbol ‘-’):  [13]

Although adjectives have been reported to reduplicate regularly in the plural in one Talodi language, Tuwal (Stevenson 1957:39, Crocker 1982), this does not happen in the four languages surveyed here, outside of the size adjectives already shown:

<32>

Adjectives can be modified by an intensifier word after the adjective in all four languages; in some cases the intensifier is derived from the adjective itself (‘good’); another intensifying strategy is partial reduplication (‘black’) which, though missing in Dagik in this survey, is now reported there too (Vanderelst 2016:147):

The intensifier words that modify adjectives are recognisable as adverbs (section 9.) because: 1. like adverbs they do not inflect for agreement in their initial consonant, 2. their initial consonants g,c,m are typical of the initial consonant classes of adverbs in these languages, 3. specific adverbs (Tocho ɪ-llɪŋ) can modify both adjectives and verb phrases.

<33>

In comparative sentences, the adjective is unmarked in all four languages and the pivot (P) and standard (S) of comparison come after the adjective (A). Acheron and Dagik both show double pivot marking, by both a pivot word (ɔkkɔ/na) and a locative prefix (nɔ-/ɪ-) on the standard of comparison (see 9.4):

<35>

In all four languages the numerals 1,2,3 agree with the noun in the initial consonant as first noted in a Talodi language by Tucker & Bryan (1966:270-288), although this persists to further numerals in particular languages. The numerals 6,7,8,9 are built from various root combinations in Lumun and Tocho as first documented in Lumun by Smits (2007a), whereas in Acheron and Dagik they are built by adding 1,2,3,4 to a base 5 after the conjunction na- as first documented in Dagik by Schadeberg (1981). However, in all four languages the numeral 9 is expressed as the sum of 5 and 4.  [14]

When we compare the root combinations in Lumun and Tocho, Lumun 6 (2-5) and 7 (2-3) do not correspond to Tocho 6 (3-3) and 7 (4-3). Instead, Tocho 6 and 7 match Tira [tic], a Heiban language also present in the Saraf Al-Jamous language cluster:

<36>

Under influence from Tira, then, Tocho deviates from other Talodi languages, in particular from the base 5 strategy found in the other two Narrow Talodi languages Dagik and Acheron, which is also reported in other Narrow Talodi languages apart from Tocho (Schadeberg 1981:45) and is thus the original system for that sub-branch. The Lumun system is different and more opaque. Lumun root compounds for 6 (-əɽa-kkʊrʊk 2-5) and 7 (-əɽa--əɽabʊrʊk 2-3) are not transparent sums or products of their component roots, despite Lumun 3 (-əɽa-bʊrʊk 2-1) being a transparent sum, so -əɽa is functionally opaque or polysemous in Lumun. In Lumun 6 (2-5), -əɽa functions as ‘successor’, thus 6 = ‘successor-5’. A successor etymology in Lumun 7 (2-3) would imply ‘successor-3’, which could be an opaque rendering of ‘successor-three[s]’ = ‘successor-6’ = 7. Another possible analysis is that 7 (2-3) is an abbreviation of 2-2-3 = 7 (Smits 2017:295). Thus, of the three numeral systems found in these four languages, we have two historic systems of opaque compounding in Lumun and the base 5 strategy in the Narrow Talodi sub-branch, plus the Tocho system which resembles neighbouring Tira instead.

<37>

Ordinal numerals are formed in various ways that appear to support definiteness of the noun they are modifying. Tocho adds a prefix a- also seen with a definite nominal function in Dagik in 8.2, Acheron uses genitive prefixation (3.2) where the relative prefix -ɛ- is used in western Acheron, and Lumun and Dagik use a relative prefix ɪ- (10.4). Special forms are used for the lowest ordinals in Acheron and Lumun, discussed below. In languages other than Tocho, the expressions for ‘second’ onwards are unexpectedly plural (‘days of four’?, ‘days that are four’?), but function as ordinal:

<38>

Acheron and Lumun share a root for ‘first’ which appears to have the etymology *nɔ-t̪ɔn ‘on-mouth’. This etymology supports an analysis with genitive prefixation /Cɔ-/ before *nɔ-t̪ɔn, which is also evident from the use of /ɛ-/ on this item in western Acheron, which appears in the genitive construction (3.2). Acheron has special forms for ‘second’ and ‘third’ as well, derived differently in the eastern and western dialects. The eastern dialect forms (m~a-m-ɛɽak, m~a-m-ət̪t̪ək) have the prefix /a-/, which is productive in Tocho ordinal numerals, with a reduplicated agreement consonant in front of it. The western dialect forms (m-ɛ~y-ɛɽak, m-ə~y-ət̪t̪ək) have a reduplicated root vowel and an internal plural prefix /y-/ that is also found in two irregular plural adjectives /b-a-b-arək, w-ɪ-y-arək/ ‘male (sg, pl)’, /b-a-b-ayɪ, w-ɪ-y-ayɪ/ ‘female (sg, pl)’.

However, all four languages also have systems of birth-order terms in families. Here, ordinal number meaning is included in the noun, so ordinal numerals are not used in this culturally important domain:

<39>

Where a series terminates, the repeated terms are marked by ‘+’. Where son and daughter terms merge, they are marked by ‘=’. Son and daughter series both terminate at the same number in each language, but this number differs in every language. Son and daughter terms merge to gender-neutral terms at the fifth-born child in the three Narrow Talodi languages Tocho, Acheron, Dagik.

<40>

Acheron and Tocho use base 10, and Tocho has an alternate suppletive term for multiples of 10:

Lumun and Dagik have base terms for both 10 and 20, where Lumun has a suppletive form for 10 added to 20 in higher multiples, and Dagik uses a pluralised phrase for multiples of 20:

<41>

Since Dagik and Acheron also use 5 as a base for 6,7,8,9, the four languages are all different in which base numerals they use, although all four languages have 10 as one of their bases - and none report a stable term for ‘hundred’:

<42>

Base numerals and other higher numerals behave like nouns in Talodi languages in a number of ways. First, as originally noted by Schadeberg (1981ː45,154), the numeral ‘five’ has a noun phrase etymology ‘one hand’, and in Dagik higher multiples are transparent noun phrases that imply similar counts of fingers and toes. Second, in Dagik, to some extent in Tocho, and on ordinal numerals in Acheron, higher numerals take genitive marking, as in Dagik mawɪ mɔ-n̪ɪpɽa ‘ten days’. Third, when a base numeral is modified by another numeral, the modifier agrees with the head numeral in the initial consonant, as in Acheron ɡʊrrʊŋ ɡ-ɛɽak ‘20’ lit. ‘two ten’. Note, however, that the Acheron the base numeral 10 does not pluralise, nor the Lumun base numeral 20 as in arrɪal w-ɛɽa lit. ‘two twenty’ – a departure from the expected behaviour of nouns (unlike Tocho which has a suppletive plural nat̪t̪ɪ ‘tens’, and Dagik where representation of multiples of 20 use the pluralised noun ‘bodies’). Fourth, higher numerals that do not take genitive marking have no agreement at all with their head noun, thus appearing as a dependent noun as in Acheron məŋki ɡʊrrʊŋ ‘ten days’, unlike the adjective-like lower numerals that agree with the noun they modify in the initial consonant. Fifth, Tocho and Lumun base numerals have an initial /a-/ which can be interpreted as a plural nominal prefix, also found in people group nouns (1. Ethnology), in some other nouns (2.1), and in the collective quantifier immediately below (5.4). Sixth, two Lumun base numeral terms are borrowed Arabic nouns: alkaɪrɛ ‘ten’ refers to the former Sudanese coin of ten piasters (Smits 2017:62), borrowed with the Arabic article al- retained, and a-rrɪal ‘twenty’ adds the plural prefix /a-/ to the Arabic monetary term riyal (Smits 2017ː296).

<43>

All four languages have a collective quantifier:

The quantifier in Dagik and Lumun behaves like some base numerals as it appears as a second noun with an initial plural prefix *a- and no agreement with the head noun. The quantifier in Tocho and Acheron is not, or is no longer, a noun because it agrees with the initial consonant of the head noun. Since reconstruction is safe when based on Dagik and Lumun (1.2), the collective quantifier can be reconstructed as a noun in the proto-Talodi ancestor language. The nominal plural *a- is preserved in all four languages, but with agreement with the head noun developing in some languages.

<45>

All four languages distinguish eight personal pronouns, as first documented by Smits (2007a) in Lumun. The eight pronouns are generated by the presence or absence of speaker, hearer and augmentation. This gives three singular pronouns, one dual pronoun (speaker and hearer), and four augmented pronouns above singular or dual number, together constituting a “minimal/augmented” pronoun system (Thomas 1955):  [15]

<46>

Vanderelst (2013) lists eight pronoun sets for Dagik for different sentence contexts. Since therefore Talodi languages are potentially rich in pronoun sets, one further pronoun set was compiled in each language with its own respective function. Once again, all the sets have eight pronouns in a minimal/augmented system:

<47>

Possessive pronouns in predicative and attributive use tend to follow the same strategies as adjectives, but there are some variations. In predicative use, the predicative pronoun appears immediately after the noun in the first three languages and only Acheron uses a copula verb, just as with adjectives. However, the Dagik predicative possessive pronouns have initial a-, also seen on ordinalised numerals in Tocho (6.2), and a plausible function uniting these occurrences is a definite nominaliser, ‘(the one that is) four’, ‘(the one that is) mine’.  [16]

<48>

In attributive possessive pronouns, a relative prefix is not used in Lumun although it occurs in Tocho and Dagik just as in attributive adjectives. Attributive and predicative possessive pronouns are thus homographs in Lumun, but in this instance a difference in tone was heard, tʊt̪t̪rʊk t-ɪ̂n ‘the pig is mine’, tʊt̪t̪rʊk t-ìn ‘my pig’. Also, the attributive possessive pronoun in Dagik is always [+ATR], which may therefore be a further attributive marker in Dagik.

<49>

In Acheron, the agreement consonant prefix is reduplicated in most possessive pronouns, and the relative prefix *ɪ- is preserved after the reduplicated agreement consonant, just as Acheron also preserves the definite nominaliser *a- after a reduplicated agreement consonant in some ordinal numerals (6.2). The relative prefix is preserved in its original high front vowel quality *ɪ-, whereas the productive relative prefix in Acheron relative clauses is now /ɛ-/ or /ə-/ (10.4).

<51>

Adverbs of pace ‘slowly’, ‘quickly’ use reduplication, also apparent in the truth adverb ‘really’ in Lumun and Dagik but obscured by phonological change in Tocho and Acheron:

The initial consonant class of a given adverb can vary from language to language, as can the first vowel which thereby also precedes the adverb root. The truth adverb ‘really, truly’ and the time adverb ‘today’ have an initial [+ATR] vowel *i in Acheron and Dagik which contrasts with the adverb-forming locative prefix*ɪ- seen in other adverbs. The word for ‘in daylight’ in all four languages contains the noun *cəŋki ‘sun’ with locative prefixation.

<52>

Place adverbs have locative *n- in the three Narrow Talodi languages Tocho, Acheron, and Dagik, as also in the time adverbs ‘at night’, ‘tomorrow’:

All four languages have a contrast between ‘up’ and ‘down’ locative prefixes on place adverbs, as seen earlier in toponyms for local ethnic groups in the three Narrow Talodi languages (1. Ethnology), and previously reported in Lumun in predicate locative expressions (Smits 2007b). The position adverb cɪk ‘down there’ (9.4) combines with place adverbs containing the down prefix t̪- in some languages as /cɪt̪-t̪-/:

<53>

Adverbs of quality are derived from adjectives using locative prefixes *ɪ- (Dagik tɪ-) or *d̪ɔ-, often with root consonant gemination, whereas in Lumun the adjective modifies lɔn ‘word’ in an adverbial noun phrase, although a similar derived adverb ɪ-kkɪ́d̪ak ‘badly’ has since been reported in Lumun as well (Smits 2017:621):

<54>

The four languages all have certain monosyllabic positional adverbs, as first documented in Lumun by Smits (2007a):

<55>

The four languages all have some further positional adverbs referring to certain frequent locations. These are derived from nouns in the same way as adverbs of quality are derived from adjectives, using a locative prefix and gemination of the first root consonant:

<56>

The locative markers *ɪ,*nɔ seen in adverbs may also be placed productively before other nouns. These are not adverbs but regularly inflected nouns, though they may be used adverbially. They differ from derived adverbs as there is no root consonant gemination: they have the usual structure of nouns with a noun class prefix before the root, and the locative marker is added in front. Locative-inflected nouns are also unlike adverbs formed by locative prefixes (including place adverbs in 9.2) in that there is variation from language to language as to whether or not the locative marker is joined to the noun, as shown in the table and discussed below:

<57>

As with genitive markers (3.2), one test for whether the locative inflections are joined to front of the noun is whether they trigger the lexical process of intervocalic voicing on the word-initial noun class consonant when it is a plosive. In Lumun, voicing of the noun class consonant suggests that the locative marker is prefixed to the noun. In Tocho, we have a distinction between /ɪ-, tɔ-/ which trigger voicing and hence are prefixed, and /ənti/ ‘from’ which does not trigger voicing and hence is separate from the noun (for /nɔ/, only additional geminated forms were elicited in Tocho, nɔ-ccʊ ‘on the rakuba’, nɔ-ppaŋ ‘with the rest’ so no conclusion could be made about /nɔ/ before nouns with noun class prefixes). In Acheron, the initial consonant is always voiced, so the intervocalic voicing test does not help. Instead, Acheron writers proposed that the locative markers are joined in some frequently-used expressions as in |ɪ.yɔk| ‘in the well’ |nɔ.bɪ.ya| ‘on the tree’, but separate when used with other words as in |ɪ|ya.zɔ| ‘in the grass’, |nɔ|ɔ.ɪŋ| ‘on/with me’. I have transcribed these examples using foot boundaries to suggest a possible prosodic difference that needs further investigation. In Dagik, the noun class consonant becomes voiced in the presence of the locative marker, but the intervocalic voicing test does not diagnose wordhood in Dagik as it is triggered at word boundaries in a number of clause contexts (sections 10.,11.). However, there is evidence that /tɪ/ is detachable because it occurs as a verbal extension on the copula /a-ɾɪ/ before a main verb (4.4).

<58>

All four languages have SVO word order (both here and throughout sections 7.-10.). The verb agrees with the subject noun in the initial consonant prefix. There is no case marking on subjects or objects.

<59>

Deviation from SVO order was found within imperative verbs, where a plural subject is marked by a verb suffix in the three Narrow Talodi languages Tocho, Acheron, and Dagik (4.1). The plural suffix (marking a plural addressee of the command) occurs before an object suffix, Acheron /ɪɟɛ-nɔŋ-ɛŋ/ (give-2pl.imp-3pl.obj) ‘give (pl) themǃ’, attesting V-s-o order within the verb, unlike Lumun where plural imperatives have been regularised to s-V-o by use of the regular 2pl subject prefix n- (4.1, 8.1). Subject marking by verb suffix is also observed in Dagik relative clauses, pat̪ɪ ɪ-p-ɔsəɡɪt̪-ɪ ‘the man who saw me’ ~ ‘the man who I saw’, where the 1sg verb suffix /-ɪ/ acts as either object or subject, distinguished by a tone difference that still needs to be described. VSO is previously reported as a variant word order in Masakin Tuwal (Stevenson 1957:38-39).

<60>

Predicate nominals use one of two copulas (with personal names lacking noun class prefixes as subject, the copula takes a default singular agreement prefix p- or b-):

Predicate locatives use various copula verbs, or no copula in Tocho:

In Tocho, the locative adverb ɪppɔŋ ‘in the near-field’ appears as a predicate that agrees with the subject noun by its agreement prefix p-. In earlier sections, we saw that predicate adjectives and predicate possessive pronouns lack a copula in Tocho, Lumun, and Dagik, hence in Tocho that pattern is extended to predicate locatives also.

<61>

The different parts of speech in this survey occur as the complements of the following copulas. In Acheron there is one copula verb for all parts of speech, but the other languages have several copulas, or lack a copula verb in some constructions, at least in present tense:

Three copulas used with adjectives, nouns, and verbs are widespread enough that they go back to the ancestor proto-Talodi language:

<62>

All four languages have a complementiser particle occurring before the complement clause. The complementiser *na in the three Narrow Talodi languages shows an irregular sound shift in Dagik *n>m also seen in ‘today’, ‘tomorrow’ (9.1).

<63>

The relative clause follows the head noun in all languages, and is marked by a relativiser prefix on its verb, although the same prefixes are also seen on attributive adjectives (5.1), genitive nouns (3.2), ordinal numerals (6.2), attributive possessive pronouns (8.2), and verbs of interrogative clauses (10.5):  [19]

<64>

The prefix *ɪ- may be reconstructed from Lumun and Dagik, reducing to schwa in eastern Acheron, although the original high front vowel quality is apparently fossilised in Acheron possessive pronouns (8.2). Relative prefixes in western Acheron and Tocho are more different, however.

The western Acheron relative prefix /-ɛ-/ differs from those in other Talodi languages in that it appears after the agreement consonant, not before it. A similar prefix /-é-/ in Moro [mor] (Jenks 2013), a larger language group bordering Acheron in the west, likewise appears after the agreement consonant rather than before it, and moreover it also occurs on genitive nouns and attributive adjectives just as the western Acheron prefix does (3.2, 5.1). Moro is a likely source for the western Acheron relative prefix because western Acheron dialect speakers report bilingualism in Moro, and Moro influence is also suggested by occurrence of a dental fricative /ð/ in western Acheron, which is frequent in Moro but atypical in the Talodi family (Norton & Alaki 2015:84).  [20]

<65>

The Tocho prefix /ɔN-/ matches the initial *ɔŋ- found on pronouns of Narrow Talodi languages (Schadeberg 1981:155). Smits (2012) finds that the Lumun cognate /ɔ́-/ on pronouns and some nouns is a marker of specificity, and its extension to relative clauses in Tocho is consistent with the fact that relative clauses make their head noun specific.

<66>

All four languages have interrogative pronouns that include extended plural forms of ‘who’. Those in Lumun, Tocho and Acheron show recurrent suffixes /-t̪a/, /-t̪ɪ/ or /-ɪ/. Verbs that follow interrogative pronouns often take additional prefixes, a- in Lumun or relative prefixes in Narrow Talodi languages, and placement of the time and place interrogatives ‘when’ and ‘where’ varies between initial and final position in the clause.

<67>

All four languages have a clause-final particle for yes/no questions. There is a polarity switch /ɪ/ ~ /a/ between the yes-no question particle vowel and the final vowel of the interrogative pronouns in Lumun and Narrow Talodi languages.

The demoted actor can be mentioned in the passive construction in the three Narrow Talodi languages, either after the preposition əŋa in Tocho and Acheron (also in 8.1) or after a different passive suffix -ŋɔ in Dagik. In Lumun, mentioning a demoted agent was only achieved by an object cleft construction that does not use the passive suffix, ʊmad̪ɔn wa [ɟɔn pɔccɔɡ-ɔ-d̪ɛ] ‘An elephant is [(what) John caught].’

The pluractional marking strategies of partial reduplication and extension also appear in habitual aspect in Tocho and Lumun (4.3).

<69>

The four languages show the same word orders on all parameters surveyed. Within the clause, there is consistent head-first word order on all of the following parameters (Dryer 1992):

The NA and PN word orders are less widely established than the rest. NA order is clear in eastern Acheron where unmarked adjectives after nouns are attributive, but in all other languages unmarked adjectives after nouns are predicates, and when they modify the noun they are marked by a relative prefix, so these are actually ambiguous with NRel word order. Nevertheless, the NA word order occurring in eastern Acheron is reported elsewhere as marginally possible in Lumun and Dagik (Smits 2007a:42, Vanderelst 2016:145), and furthermore numeral adjectives and demonstrative adjectives follow the noun in all four languages.

<70>

The question of whether Talodi languages have true prepositions before nouns in the order PN is not a trivial one. Genitive markers (3.2) and locative markers (9.4) come before the noun in the four languages, but they may or may not be separate from their nouns, and could be case modifiers rather than prepositional heads, in a given language. However, credible prepositions of accompaniment are seen in Acheron and Tocho with a pronoun complement and, in Acheron, governing the selection of different pronoun sets (8.1). Pivot words in Acheron and Dagik comparative sentences before locative-marked nouns are also candidate prepositions (5.4).  [21]

<71>

Turning to word orders at the clause level, clause particles do not all occur in head-first position before their clause. In all four languages, complementisers and relativisers come before their clause (head-first), but yes/no question particles come after their clause:

<72>

Reviewers of this paper point out that the solitary exception of question particle order calls into question whether the question particle is actually a head (see also Dryer 1992:102-123). It thus recalls the negative verbal modifier particle *tɪ found in some Narrow Talodi languages (4.6). There are no other exceptions to head-first word order in the survey, nor in published analysis of lengthy Lumun texts (Smits 2007a, Stirtz 2012), although since this survey was done Dagik has been reported to allow OV order for object focus (Vanderelst 2016:233). Another language, Lafofa [laf], classified as a Talodi language by Greenberg (1963) and Schadeberg (1981), shows multiple opposite word orders OV, AN, PossN, NP (Stevenson 1957:43-46, Tucker & Bryan 1966:270-288), but its membership of the Talodi family is disputed on independent lexical grounds by Stevenson (1964), Blench (2013), and Norton & Alaki (2015).

<73>

The four languages surveyed all share the following grammar features. As the survey builds on previous work on Talodi languages starting with Stevenson (1957:35-42), this list provides the stable features that jointly characterise the Talodi languages in the current state of knowledge:  [22]

No grammatical category surveyed here was found to be marked by a minimal tone contrast on written homographs in all four languages. However, tone is known to have a role in forming imperative verbs in both Lumun and Dagik (Smits 2017:339-341, Vanderelst 2016:108), and since Lumun and Dagik represent the two branches of Talodi, imperative tone goes back to proto-Talodi and is therefore likely to be inherited by further Talodi languages. In any Talodi language shown to have imperative tone, it must be mastered by readers wishing to use instructional texts as well as narrative texts.

<74>

The four languages Lumun, Tocho, Acheron and Dagik were confirmed as separate languages on lexical grounds in Norton & Alaki (2015). Grammatically, they are all found to be different from each other in the numerals they use as bases and in the sizes of their birth-order noun sets. Otherwise, repetition of grammatical properties across at least some of the languages is the norm.

The eastern and western dialects of Acheron (Norton 1995) are found to differ grammatically here. They have different relative prefixes, of which only the western prefix is found on attributive adjectives and genitive nouns. The dialects also present different derivations of the ordinal numerals ‘second’ and ‘third’.

The three languages spoken on adjacent lands at Saraf Al-Jamous, Lumun, Tocho, and Acheron, which were previously classified together as a group (Stevenson 1956:102), share only a few properties in this survey that distinguish them from Dagik located further away to the south:

However, the three Narrow Talodi languages Tocho, Acheron, Dagik share many more properties that distinguish them from Lumun: (here PNT = Proto-Narrow-Talodi)

The weight of the evidence from this grammar survey therefore supports the proposal of Norton & Alaki (2015) that Lumun belongs in a separate branch from Tocho, Acheron, and Dagik and has evolved separately from them over a longer period.

References

Alaki, Thomas Kuku 2008

‘Tira numerals’. Numeral Systems of the World’s Languages Project. http://lingweb.eva.mpg.de/numeral/Tira.html

Alamin, Suzan 2003

An initial syntactic description of the Asheron language, a language of the Nuba Mountains. MA Dissertation (University of Khartoum)

Bhat, D.N.S. 1999

The Prominence of Tense, Aspect, and Mood. Amsterdam: Benjamins

Bird, Steven 1999

‘When marking tone reduces fluency: An orthography experiment in Cameroon’. Language and Speech 42:83-115

Blench, Roger 2013

‘Splitting up Kordofanian’. In Schadeberg, Thilo and Roger Blench (eds.), Nuba Mountain Language Studies, p.571-586. Cologne: Rüdiger Köppe

Crocker, Ellen 1982

‘Grammar write-up [of Masakin Tuwal]’. Unpublished manuscript

Crocker, Ellen 1984

‘Some vocabluary from Masakin Tuwal’. Unpublished manuscript

Dryer, Matthew 1992

‘The Greenbergian word order correlations’. Language 68:81-138

Greenberg, Joseph 1963

The Languages of Africa . The Hague: Mouton

Jenks, Peter 2013

‘Noun phrases in Moro’. In Schadeberg, Thilo and Roger Blench (eds.), Nuba Mountain Language Studies, pp.57-82. Cologne: Rüdiger Köppe

Kröger, Oliver 2012

‘Typology put to practical use: A participatory approach to initial grammar research’. In Brenzinger, Matthias and Anne-Marie Fehn (eds.), Proceedings of the 6th World Congress of African Linguistics, Cologne, 17-21 August 2009, pp.155-168. Cologne: Rüdiger Köppe

Kutsch Lojenga, Constance 2014a

‘Orthography and tone: a tone-system typology with implications for orthography development’. In Michael Cahill & Keren Rice (eds.), Developing Orthographies for Unwritten Languages, pp.49-72. Dallas: SIL International

Kutsch Lojenga, Constance 2014b

‘Basic principles for establishing word boundaries’. In Michael Cahill & Keren Rice (eds.), Developing Orthographies for Unwritten Languages, pp.73-106. Dallas: SIL International

Munro, Pamela 2007

‘From parts of speech to the grammar’. In Thomas Payne & David Weber (eds.), Perspectives on Grammar Writing, pp.71-111. Amsterdam: Benjamins

Norton, Russell 1995

Variation and change in the phonology of Asheron. MA Dissertation (University of Essex)

Norton, Russell 2000

‘The noun classes of Asheron’. Occasional Papers in the study of Sudanese Languages 8:23-55

Norton, Russell 2012

‘Non-concatenative morphology in Dinka and Arabic: Augmentation and replacement’. In Matthias Brenzinger and Anne-Maria Fehn (eds.), Proceedings of the 6^th World Congress of African Linguistics, Cologne, 17-21 August 2009, pp.169-180. Cologne: Rüdiger Köppe

Norton, Russell and Thomas Kuku Alaki 2015

‘The Talodi languages: A comparative-historical analysis’. Occasional Papers in the study of Sudanese Languages 11:47-162

Norton, Russell and Thomas Kuku Alaki 2017

‘Torona: A disappearing Talodi language of Sudan’. In Gratien Atindogbe and Evelyn Fogwe Chibaka (eds.), Proceedings of the 7^th World Congress of African Linguistics, Buea, 17-21 August 2012, pp.152-177. Bamenda: Langaa

Payne, Huddleston and Pullum 2010

‘The distribution and category status of adjectives and adverbs’. Word Structure 3(1):31-81

Roberts, David 2010

‘Exploring written ambiguities can help assess where to mark tone’. Writing Systems Research 2(1):25-40

Schadeberg, Thilo 1981

A Survey of Kordofanian, Volume 2: The Talodi Group. Hamburg: Helmut Buske

Smits, Heleen 2007a

Noun and noun phrase in Lumun (Kordofanian). MA Dissertation (University of Leiden)

Smits, Heleen 2007b

‘Notes on the description of topological scenes in Lumun’. Unpublished manuscript (University of Leiden)

Smits, Heleen 2012

‘The prefix /ɔ́-/ in Lumun kinship terms and personal names’. Occasional Papers in the study of Sudanese Languages 10:95-114

Smits, Heleen 2013

‘The locative-applicative suffix in Lumun’. In Schadeberg, Thilo and Roger Blench (eds.), Nuba Mountain Language Studies, pp.219-236. Cologne: Rüdiger Köppe

Smits, Heleen 2017

A Grammar of Lumun: A Kordofanian Language of Sudan. Leiden: Netherlands Graduate School of Linguistics

Snider, Keith 2014

‘Orthography and phonological depth’. In Michael Cahill & Keren Rice (eds.), Developing Orthographies for Unwritten Languages, pp.27-48. Dallas: SIL International

Stevenson, Roland 1956-57

‘A survey of the phonetics and grammatical structure of the Nuba Mountain languages’. Afrika und Übersee 40:73-84,93-116, 41:27-65,117-152,171-196

Stevenson, Roland 1962-64

‘Linguistic research in the Nuba Mountains’. Sudan Notes & Records 43:118-130, 45:79-102

Stirtz, Timothy 2012

‘Lumun participant reference in narrative discourse’. Occasional Papers in the study of Sudanese Languages 10:115-140

Thomas, David 1955

‘Three analyses of the Ilocano pronoun system’. Word 11:204-208

Tucker, A.N. & Margaret Bryan 1966

The Non-Bantu Languages of North-Eastern Africa . London: Oxford University Press

Vanderelst, John 2013

‘Personal pronouns in Dagik (Talodi, Kordofanian)’. In Schadeberg, Thilo and Roger Blench (eds.), Nuba Mountain Language Studies, pp.157-176. Cologne: Rüdiger Köppe

Vanderelst, John 2016

A Grammar of Dagik: A Kordofanian Language of Sudan. Cologne: Rüdiger Köppe