A Brief Introduction to the Turi Language of Eastern India

2.1 Phonology

Table 1 presents the consonant phonemes in Turi which we have been able to identify. All four affricates, all consonants with breathy voice, /ɳ/ and /ʋ/ are restricted in our data to loan words from Indo-Aryan. For ease of presentation, in all Turi example sentences aspiration and breathy voice are represented by <h>. In Table 1 and in the word list in Appendix 1, however, these are given in standard IPA, with superscripted aspiration [tab. 1].

Munda languages are known for their pre-glottalized, voiced stops, often with a nasal release, i.e. [^ʔb] or [^ʔb˺^m]. This class is also found in syllable-final position in Turi, although there is considerable variation which requires further study. Cf. e.g. from the word-list in Kobayashi et al. (2003, 353, 359) words for ‘hair’ (#2) from Mundari such as [u:b˺]/[u:b] or Santali [u:b˺]/[u:p˺], with Turi [uʔm], or ‘eye’ (#5): Ho [meɖ˺], Mundari [med˺]/[med]/[meːʤ] and Santali [meːt˺]/[met˺]/[meˑt˺] (Kobayashi et al. 2003, 347, 353, 359) with Turi [mɛʔn]. As can be seen, Turi tends to lose the stop entirely and retain the nasal – and often also glottalization – although alternative forms are also found in the texts, e.g., [dub], [dubʔm] and [duʔm] ‘sit down’.

Table 2 presents the monophthong phoneme inventory of Turi as found in our database. The status of /ə/ as a phoneme is unclear. At present our data suggest that it is best viewed as an allophone of /ɑ/. All vowels can be nasalized; it is not yet clear if nasalization can be phonemic. Vowel length, however, is not phonemic, and all vowels other than /ə/ can be realized as long in certain environments, such as word-final positions [tab. 2].

The following diphthongs are also found in our corpus in native words: ə̯ɛ, i̯a.

2.2 Nouns

There is no grammatical gender in Turi. However, object indexing on the verb is sensitive to the animate/inanimate opposition as nouns with animate reference trigger verbal agreement in the 3rd persons, while nouns with inanimate reference do not.⁸ Nouns may appear in the singular, which is morphologically unmarked, in the dual, marked by =kin, or in the plural, marked by =kun.

The morphologically unmarked noun – especially if it has inanimate reference – can also have plural reference if this is clear from context or is indicated elsewhere in the text, e.g. sɔbu ʈɔpi ‘all the hats’ in (22) or ʈɔpi ‘the hats’ in (23).

2.3 Pronouns

Independent pronouns are found in the singular, dual and plural. Dual forms can be reinforced by adding the form baran=kin ‘both=du’, e.g., aliŋ baraŋ=kin ‘we both (excl)’, and plural pronominal forms of the 1st and 2nd persons can be pleonastically marked for =ku ‘pl’, e.g. alɛ=ku ‘we (pl.excl)’. Table 3 presents the independent pronouns [tab 3].

2.4 Numerals

In Odisha Turi, only the numerals one-three are of Munda origin; beginning with ‘four’, Indo-Aryan loan words are used. Grierson (1906, 128) also cites the form pūniā ‘four’, which is not in use in Odisha Turi. These are given in Table 4 [tab. 4].

miaʔn ‘one’ is also commonly used as an indefinite article; cf. (2).

The morph =gɔɽ, which is homophonous with a classifier in many neighbouring Indo-Aryan and Munda languages of the region, is also found in our texts, together with the Indo-Aryan numerals ʧar ‘four’ and panʧ ‘five’, but not with the lower numerals of Turi origin. However, it does not appear to be a classifier in Turi, as it is used in counting in our corpus (cf. (3)), but not with a following noun; cf., e.g., miaʔn duba ‘a tree’ in (2) or miaʔn phɛriʋala ‘a hawker’ in example (40). We therefore tentatively gloss it as “count”, due to its counting function.

2.5 Case

Odiya Turi has five cases:

• Unmarked – The unmarked (or zero-marked) case marks subjects (e.g., maʤur ‘the peacock’ in (9)) and inanimate objects (e.g., sɔbu ʈɔpi ‘all the hats’ in (22)).

• Objective – This case is used with primary objects with verbs of speech, cf. (4), and animate secondary objects (cf. uni kɔɽa=kɛ ‘that boy’ in example (13)).

=kɛ can also mark definite inanimate secondary objects (or P, in typological terms), as in (5), although this is not obligatory, as sɔbu ʈɔpi ‘all the hats’ in (22), mentioned above, shows.

• Genitive – The genitive is used to incorporate a nominal phrase into a larger nominal phrase. There are two genitive markers: =aʔ/=aʔa, used with pronouns (6), and =rɛn, used elsewhere; cf. (7).

With adnominal inalienable possession involving kinship terms, a different construction is used. Here, the noun is followed by the possessive marker -ta which is followed by an enclitic form of the independent pronoun, as with the two forms of ba ‘father’ in (8).

• Locative – The locative marks the location of an entity. It has the form =rɛ (cf. (9)).

• Instrumental – The instrumental case marks the instrument with which an action is carried out or the cause of an action. It has the form =tɛ, as shown in (10).

Case markers follow number markers in Turi; cf. the form bandra=kun=kɛ [monkey=pl=obl] ‘to the monkeys’ in (4) above or ʧhaʋa=kun=rɛn [child=pl=gen] ‘of the children’ below in (23).

2.6 General Introduction – Lexical Verbs

The Turi verb system shows many similarities to those of other North Munda languages, although with some differences, as we show in the following. Figure 1 gives a somewhat simplified structural overview of the affirmative finite verb in Turi and Figure 2 that of the negative verb [figs 1-2].

Figure 1 The basic schema of the affirmative finite verb in Turi

Figure 2 The basic schema of the negative finite verb in Turi

The first element of the predicate is the lexical base, which usually consists of a single lexical morpheme. This is followed by portmanteau TAM/basic voice markers (= active or middle), object indexing for objects with animate reference, the finite marker /a/, and subject indexing for subjects with animate reference.

Table 5 gives an overview of the enclitic subject/object markers on the verb. With the exception of the 3rd person singular, these are all highly similar to the full forms of the pronouns given in Table 3. The two forms of the 1st person singular are speaker-specific free variants, while in the 3rd person singular in Odisha Turi =i indexes an object and =ɛ indexes a subject at the respective positions shown in Figures 1 and 2. In contrast, in Jharkhandi Turi =i indexes both subject and object. Otherwise, all indices in Table 5 can index both subjects and objects, at their respective positions within the predicate [tab. 5].

The finite marker /ɑ/ is realized as [ɑ] before subject markers beginning with a consonant but is elided before the 1st person singular =ɛŋ/=iŋ. Before the 3rd person singular marker =ɛ, the finite marker is raised and realized as /ə/. These two vowels are pronounced together as the diphthong [ə̯ɛ] Cf. the respective forms in (11).

The P-argument, roughly corresponding to the “direct object” with mono-transitive verbs, is indexed on the predicate before the finite marker when the reference of this argument is animate. This is shown in (12), from Jharkhandi Turi, where =i precedes the final finite marker =a.

(Jharkhandi Turi)

When the finite marker is not overtly realized because the subject is the 1st person singular, the object-index directly precedes the subject index, as in (13). Note that these two indices are pronounced as two separate syllables in (13), from Odisha Turi.

Table 6 presents the TAM/basic voice categories that we have identified in Odisha Turi. As the Odisha Turi data presently do not contain examples involving the active perfect, the active present-perfect marker in Table 6 is from Grierson’s (1906, 130) data for the Ranchi dialect of Turi [tab. 6].

The alternate forms of the past perfect marker -laʔ/-lɛʔ and the middle past marker -kɛn/-ɛn appear to be speaker-specific (cf. (14)-(15)).

In a few isolated forms (cf. e.g. (16)), the middle marker -ɔʔ, or its shortened form -ɔ, appears before the middle-voice past marker -ɛn⁹ with no apparent semantic distinction to the use of -ɛn alone. After the short form /ɔ/, the /ɛ/ of -ɛn is elided and the form is realized as [ɔn]. The decline of the use of the middle markers here is perhaps connected to the fact that these forms are still recognizable as middle-voice past forms, distinct from the active forms, even without the marker -ɔ(ʔ).

With respect to the active past forms: In our data, -ɛkɛn is a general past imperfective used for prolonged actions such as progressives, states and iterative actions. In contrast, -tad appears to be restricted to single actions; we analyse it as a pure past tense, not as a perfective past tense.

This distinction in the active past seems to hold not only for Odisha Turi but also for the data from Grierson for Ranchi, Jashpur and Sarangarth Turi, at least tendentially. Compare the use of - ɛkɛn in (17) from Odisha Turi for a prolonged action with that for -tad below in (22), also from Odisha Turi, or in (18), from Jashpur Turi, from Grierson.

(Jashpur Turi, Grierson 1906, 131)

Standard negation in Turi is marked by ka, which appears directly before the predicate. The subject index attaches directly to this marker instead of appearing as the final enclitic on the predicate, where it is found in non-negated clauses (cf. again Figures 1 and 2). When followed by the 3rd person singular animate marker =ɛ, underlying /kɑ/ is realized as [kə].

Only one object can be marked on the verb in Turi, and only if this object is considered animate. Object indexing in Odisha Turi follows a secundative alignment pattern, i.e., a distinction is made between the indexing of primary objects (= Patients and Goals) and secondary objects (= Themes), as opposed to indirective alignment, which distinguishes between direct and indirect objects.

A note is necessary here with respect to terminology. With (mono-)transitive verbs, the primary object is P (for ‘Patient’), corresponding roughly to the direct object of English. Bitransitive predicates, on the other hand, have G (for the ‘Goal’ of the action) and T (for ‘Theme’), i.e., that entity which ‘moves’ to the goal with verbs such as give. When P and T receive the same marking, e.g., the accusative case in Latin or Sanskrit, they together form the category of direct objects. G is then marked differently from P and T, e.g., the dative in Latin or Sanskrit, and is the indirect object.

In Odisha Turi, P is indexed on the predicate when it references an animate entity, as in (12)-(13) above, repeated here for convenience as (20)-(21).

However, an animate G is also indexed on the predicate, as (22) shows, without any applicative to raise it to object status. In (22) the =i of the form ɛm-tad=i=a=ku refers to the man to whom the monkeys gave the hats (not explicitly mentioned in (22)). As P and G are similarly indexed at the same slot on the Turi verb but not T, Turi has a primary/secondary object distinction. The secondary object is then T, the NP sɔbu ʈɔpi ‘all [the] hats’, which is not indexed on the verb.

The gentive-marked possessor of G can also be indexed as the primary object, as with ʧhaʋa=kun ‘the children’ in (23), who are the possessors of lahiʔɲ=kɛ ‘bellies.’

As examples (24)-(25) from Grierson’s data for Ranchi Turi and Sarangarh Turi show, an applicative construction is used in other dialects of Turi when G is indexed on the verb, i.e., G is raised here to the status of an object. In contrast, in Odisha Turi this is the default indexing pattern, and no applicative is found.

(Ranchi Turi, Grierson 1906, 130)

(Sarangarh Turi, Grierson 1906, 133)

3.1 The Position of Turi Within the Munda Family – Previous Discussions

The Munda languages, to which Turi belongs, form the western-most branch of the Austro-Asiatic language family, which stretches from Central India in the west to Vietnam in the east. Map 2, from Sidwell (2015, 144), provides an overview of the spread of this family and its main branches [map 2].

Figure 3 presents the traditional internal classification of Munda, from Zide (1969, 412) [fig. 3]. In this classification, the Munda group bifurcates into North Munda and South Munda branches. The northern branch then bifurcates into Korku, spoken in central India, and Kherwarian, spoken in eastern India. Only the Kherwarian branch will be discussed in the following.¹²

Turi belongs to the Kherwarian group and is largely mutually intelligible with the languages of both the Santali and the Mundari-Ho branches, but what has remained unclear until now is to which of these two branches it belongs, or whether it possibly constitutes a third, independent branch.

Grierson (1906, 128) writes the following of Turi: “The Birhâṛ dialect is closely related to Muṇḍārī, and the speech of the Tūrīs also agrees with that language in most essential points. In a few characteristics, however, it follows Santālī, as against Muṇḍārī”. Further on the same page, he writes that “[i]n Sambalpur the Tūrī dialect is almost pure Muṇḍārī” but then goes on to note similarities with Santali, most notably with respect to phonology, which is central to the comparative method: “Forms such as pēā, three; pūniā, four, in Tūrī agree with Santālī, as does the phonology of the dialect in most points”.

Munda (1968, 46-7) on the other hand notes a number of similarities between Turi and both Mundari and Santali.¹³ This indeterminate status of Turi within Kherwarian persists until today, e.g. with the Glottolog (Hammarström et al. 2022) classifying Turi as a Mundaric language while the Ethnologue (Eberhard et al. 2023) groups it together with Santali.

3.2 The Data

In order to determine the position of Turi within the Kherwarian languages, the data for all 274 lexical items contained in Kobayashi et al. (2003, 347-67) for three dialects of Ho (Chaibasa, Goilkera and Ghatshila), four dialects of Mundari (Darigatu, Bandugara, Salgadih and Kera) as well as five dialects of Santali (Kadma, Heben, Tikahara, Hatsara and Simoldohi) were entered by Luna Hemmerling, a student aide at Kiel University, into a spreadsheet. Our own data on Odisha Turi were then added to this spreadsheet.¹⁴

Our original aim was to enter detailed data on several dialects of Turi into this file for comparison with these other Kherwarian varieties. Unfortunately, this turned out to be more difficult than anticipated. First, we have not been able to locate any speakers in the state of Jharkhand under the age of 60, and many of the speakers that we have located are either semi-speakers or have not used the language in many years. The situation with COVID-19 further complicated our work, making it impossible to reach many villages during this time, and a number of these older speakers who we had contacted before the pandemic tragically succumbed to this illness during the second wave in 2021. We have however located a relatively large number of Turi speakers from northwestern Odisha, from communities in which Turi is still being learned by the younger generation. As Covid until very recently continued to make travel to these villages difficult, a group of five Turi adults, three male and two female, were invited to Ranchi to cooperate with the authors of this study during the course of five days to elicit stories and songs and to complete as much of this vocabulary list as possible.

As the members of this group came from different villages, there is unfortunately no one single local variety which we could record but five slightly varying varieties, although all speakers were from a relatively small geographical area and their data do not differ substantially. As a result, however, our data are not as fine-grained as that for the 12 Ho, Mundari and Santali dialects documented in Kobayashi et al. (2003). Our data for Odisha Turi were therefore entered as one form of Turi, although we have included all variant forms from these respondents in the spreadsheet. Altogether, data for 227 of the 274 items in the above-mentioned list were obtained in this manner. Data from the considerably more different Jharkhandi Turi were not included, as we still have no data for many lexical entries.¹⁵

The data were then cleaned to ensure maximal comparability. First, as it is imperative that only cognate forms are being compared, we removed all loan words from the 13 varieties in the data which we were able to determine, including very old loan words from Indo-Aryan such as daru ‘tree’ (no. 118) but of course also more recent loans, such as various words for ‘lake’ or ‘sea’ (nos. 140‑1), cf. e.g. sagar, samud (from Sanskrit), dɔrɛja (cf. Persian dæɾja), etc. We also removed compounds from the list as the monomorphemic words for the relevant lexemes were already contained elsewhere in the list. E.g., transparent compounds such as e.g. ‘tear’ (no. 6), consisting of morphemes for ‘eye’ and ‘water’, i.e., ‘eye-water’, were removed from the data, since these two lexemes are already contained in the list (no. 5 ‘eye’ and no. 143 ‘water’).

Complex forms where one part was cognate with forms in other dialects in the list were handled differently; here we retained the cognate lexeme from the respective complex form and removed the other element, which was either opaque, a grammatical morpheme (e.g. a nominalizer) or a lexical morpheme found elsewhere in the list. To give an example, no. 39 ‘sweat’: In the Santali data we find the forms ud’gɛr (Heben) and ud’gər (Kadma, Tikahara and Simoldohi), but ud’gər daˑk˺ for the Hatsara form. Here, daˑk˺ is clearly the morpheme for ‘water’ and was thus removed, as it is found elsewhere in the list (no. 143), while the first element, ud’gər, was retained, as it is cognate with the other four entries.

Our goal was primarily to compare phonological developments in these languages in order to determine their genealogical relationships with one another, not lexical similarity per se. For this reason, we only consider phonetic similarity in the following. We also deleted a number of problematic entries, such as deictic units which consisted primarily of grammatical morphemes where it was not always clear whether the forms were cognate (e.g., nos. 257-63). These will have to be studied in more detail in a future study.

Although perhaps the most characteristic phonological trait of Munda languages, the preglottalization and non-audible release of syllable-final voiced plosives (e.g., [^ʔb˺^m]) were removed from the data, as these forms are often realized in Turi and many other Munda languages both as preglottalized and as non-glottalized variants even by the same speaker in natural speech, and we felt that it could skew the data if e.g. in some languages both forms occurred but only one of the two happened to be documented. For this reason, only the non-preglottalized forms were used. Finally, as no studies have yet been carried out on lexical accent in Turi or how this is to be defined, all primary and secondary accents were removed from the data for all varieties before comparison. We then deleted all lexical entries for which less than five forms in total were present from the 13 different linguistic varieties.

Finally, all entries were removed which did not have a corresponding non-Indo-Aryan entry for Turi. Altogether, these measures combined to reduce the number of lexical entries considerably, from the original 274 items – of which we have 227 for Turi – to 95. It is this smaller group that forms the basis for our comparison.

We stress here that this is meant only as a preliminary attempt to determine Turi’s position within Kherwarian, based on our current knowledge. As finer-grained data for Turi and other Kherwarian languages are added to this lexical database, our understanding of Turi and the other languages in this group will continue to improve. But as we now have enough data for a first analysis of Turi, we believe that an introduction to the language such as the present one should at least offer a preliminary discussion of Turi’s place in Kherwarian, however tentative that may be.

3.3 Results

The data were then analysed with the software COG by the Summer Institute of Linguistics. COG was chosen for a number of reasons: First, the software is openly available and easy to use, even when the respective researcher does not have extensive computational skills. As such, our results can be reduplicated by other researchers regardless of their computational training. Second, COG allows researchers to view the results in various formats, such as dendograms, (non-rooted) trees, networks, and also provides a similarity matrix for the different varieties. Third, COG allows the researcher to actively participate in the analysis, e.g., by correcting false analyses where the algorithm mistakenly views different forms as cognate which the researcher does not consider cognate, or conversely by marking these forms as cognate when the algorithm has not analysed them as such.¹⁶

Finally, COG provides analyses with respect to lexical and phonetic similarity in both a UPGMA and a Neighbour-Joining (NJ) analysis.¹⁷ Unfortunately, these are the only two algorithms offered in COG, although in view of the preliminary nature of our investigation, this will be sufficient for an initial perspective. The results of this comparison are presented in Figures 4-5.

In both Figures 4 and 5, Santali dialects all group together, all Mundari dialects other than Kera Mundari (to which we will return) group together, and all Ho dialects group together. Furthermore, the Mundari and Ho branches – together with Kera Mundari – form a branch distinct from Santali [figs 4-5]. While this classification is not surprising, it does confirm that both analyses come to very similar conclusions. In fact, they are identical with the exception of Kadma Santali and Heben Santali: in Figure 4, Heben forms the outermost branch of Santali, followed by Kadma and then the remaining three varieties (Tikahara, Hatsara and Simoldohi), while in Figure 5 Kadma and Heben together form a sister branch to the remaining three Santali dialects, which have the same internal relationships. Otherwise the two figures are identical with respect to their internal branching.

With respect to Turi, note that in both Figures 4 and 5 Turi is a sister language to the entire Santali branch, i.e., in neither analysis is Turi more closely related to Mundari-Ho than to Santali. This is more in line with the classification in the Ethnologue (Eberhard et al. 2023) than with the Glottolog (Hammarström et al. 2022), which classifies Turi as a member of the Mundaric branch.

Figures 4 and 5 also incidentally show that both algorithms classify Kera Mundari as a sister language to the entire Mundari-Ho branch, and not within the Mundari group. While this position of Kera Mundari may at first glance seem somewhat unexpected, it becomes more understandable when the history of this dialect is taken into consideration: Kera Mundari is known to be the result of a language shift by speakers of Kurux (Dravidian) to Mundari, as the result of which it “has unique characteristics and constitutes a distinct regional as well as ethnic dialect” (Kobayashi, Murmu 2008, 165). This shift provides a likely explanation for its relatively distant relation to the other members of this group, although it is considered a dialect of Mundari by the speakers themselves.