Grammatical tone

Tonal alternations triggered by grammatical elements anteposed to nouns

Some of the H-toned grammatical elements anteposed to nouns, despite being written as if they were separate words, interact with the noun they precede in a specific way, which shows that they must be analysed as bound elements. Contrary to what occurs at ordinary word boundaries when the first word ends with a H tone, they never trigger downstep, and their H tone may spread to more than one syllable.

This can be illustrated by comparing the tonal properties of ka ‘with (instrumental)’ and le ‘with (accompaniment)’. As illustrated in (1) and (2), ka has the phonological properties of a word: if the word that follows ka begins with more than two successive L-toned syllables (here: lobone [lʊ̀bɔ̀nɛ̀] ‘lamp’), the spreading of its H tone is limited to the first syllable of the following word, and a downstep occurs if the word that follows it begins with two successive H-toned syllables (here: tshipi [tsʰípí] ‘iron’). By contrast, as illustrated in (3) to (4), in the same conditions, with le, the spreading of the H tone may extend to the second syllable of the following word (here: molemi [mʊ̀lɩ̀mì] ‘farmer’), and no downstep occurs when the following word begins with two H-toned syllables (here: kgosi [qʰósí] ‘chief’).

The special tonal behaviour illustrated in (3) en (4) with le is also found with the linker introducing nouns in the function of adnominal possessors, for example tsa in example (5) and (6).

Tonal modifications of nouns triggered by the locative suffix -ng

With some of the nouns whose basic tonal melody ends with two successive L tones, the attachment of the locative suffix ‑ng may trigger the replacement of the final L tone of the noun by a H tone. However, this change is optional. For example, sedibeng (locative form of sediba ‘well’) may be pronounced [sɩ̀dìbèŋ̀] or [sɩ̀dìbéŋ̀] Note that the locative suffix itself is invariably L-toned.

Tonal modifications of nouns in 'noun + modifier' constructions

When followed by certain modifiers, the nouns whose basic tonal melody ends with two H-toned syllables undergo a tonal process by which their final H tone is replaced by a L tone. In (8), we observe that the insertion of the modifier botlhe ‘all’ does not trigger any change in the tonal melody of basadi ‘women’, whereas in (9), the insertion of the modifier ba motse ‘of the village’ triggers the replacement of the final H tone of basadi by a L tone.

This tonal change occurs whenever a word whose basic tonal melody ends with two successive H tones is immediately followed by a modifier belonging to one of the following types:

• adnominal possessor (as in (9) above),
• adjective,
• relative clause,
• demonstrative,
• the interrogative determiner -fe ‘which?’
• the negative determiner -pe ‘any, no’;
• the determiner -sele ‘other’.

The two tonal classes of verbs

Setswana verbs divide into two tonal classes: those that may have forms including no H-toned syllable (lexically toneless verbs), and those that have at least one H-toned syllable in all their forms (lexically H-toned verbs). (See the topic on verbal domains.) If no tonal process resulting in the introduction of additional H tones occurs, lexically toneless verbs surface with an entirely L melody, whereas the melody of lexically H-toned verbs can be described by positing a H tone inherently attached to the first syllable of the stem and a spreading rule that may copy this H tone on the second and third syllables of stems including two or more syllables.

This can be observed in the infinitive, if no additional element introducing a H tone is inserted between the infinitive marker go and the verb stem. Note that, in the case of lexically H-toned verbs whose stem does not include more than three syllables, the tonal melody indicated in Table 1 and 2 is that observed in sentence-internal or sentence-initial position. In sentence-final position, or when the infinitive is quoted in isolation, it is modified according to the general rule that operates in this context. For example, go rekisa ‘to sell’ has the basic tonal melody [ _ ¯ ¯ ¯ ], but is quoted as [ _ ¯ \ _ ].

The tonal morphology of verbs: general remarks

The tonal morphology of verbs (i.e., the set of rules that determine the tonal melody of verb forms) is a particularly complex aspect of Setswana grammar. Regardless of the length of their stem, Setswana verbs divide into two tonal classes, but certain tenses are characterised by the addition of a grammatical H tone attached to the second syllable of the stem. Moreover, the verb stem interacts tonally with various grammatical elements that precede it, which are written as separate words in the traditional orthography but are in fact bound to the verb stem, with which they interact in a very specific way. Finally, the tonal behaviour of verb finals is different from one tense to another.

Given the complexity of this system, it cannot be presented in detail here. In this remainder of this section, I limit myself to some relatively straightforward illustrations of the various types of tonal processes that determine the tonal melody of verb forms. But before illustrating the various types of tonal processes operating within the verb forms of Setswana, a short note about the involvement of tone in the distinction between conjoint and disjoint verb forms is in order, since this is a major source of difficulties in the analysis of the tonal morphology of Setswana verbs.

Tone and the distinction between conjoint and disjoint verb forms

The distinction between conjoint and disjoint verb forms is common among Eastern and Southern Bantu languages, but a striking particularity of Sotho-Tswana languages (including Setswana) is the importance of tonal morphology in the expression of this distinction.

In Setswana and other Southern Bantu languages, the distinction between conjoint and disjoint verb forms can be defined as follows. Conjoint verb forms can only occur in non-final position and must be immediately followed by a phrase adding some precision about the event to which the verb refers, whereas disjoint verb forms occur mainly in final position, and can only occur in non-final position if the phrase following them refers to already mentioned information.

In Setswana, the only tense in which this distinction is obvious is the present positive, in which the disjoint form is distinguished from the corresponding conjoint form by the presence of the disjoint marker a immediately after the subject index. For example, in (10), the deletion of the disjoint marker a makes the sentence ungrammatical, unless something is added after the verb, whereas in (11), it is impossible to add the disjoint marker a without deleting the adverb sentle at the same time.

As illustrated in (12), the disjoint form is possible in non-final position if followed by a le-phrase resuming the subject, le being interpreted as ‘too’, whereas le-phrases expressing accompaniment trigger the use of the disjoint form.

In this context, we observe that the conjoint and the disjoint forms differ also in the tone of their final syllable: H in the disjoint form, L in the conjoint form.

In other tenses, similar distinctions can be observed, but their manifestation is purely tonal. For example, the sentence Balemi ba tsamaile le bone has two possible readings, one in which bone resumes balemi ‘The farmers too have gone’, and another one in which bone refers to a distinct group of people ‘The farmers have gone with them’, but in the oral language there is no ambiguity, since the two possible readings are distinguished by the tonal melody of the verb.

The explanation is that ba tsamaile with the tonal melody H H L L L is a disjoint verb form, i.e., a form occurring in the same contexts as the a-form of the present positive, whereas ba tsamaile with the tonal melody H L H H L is a conjoint verb form, i.e., a form implying that the verb is followed by a phrase adding some precision to the event described by the verb.

The grammatical H tone

In some tenses, the tonal melody of the verb stem results from the addition of a grammatical H tone. In stems including two or more syllables, this H tone attaches to the second syllable of the stem and spreads up to the last syllable or to the penultimate syllable of the stem, depending on the tonal properties of the final (see the section on the tonal properties of final verbs).

The conjoint form of the present negative, illustrated in Table 3, is one of the tenses in which the grammatical H tone is inserted, and in which it invariably spreads up to the last syllable of the stem.

The tonal properties of the grammatical elements anteposed to the verb stem

The grammatical elements anteposed to the verb stem interact tonally between themselves and with the verbal stem in a specific way. In (16) and (17) we observe that, if a L-toned object index (for example, the object index mo of class 1) is inserted in one of the infinitive forms illustrated in Tables 1 and 2, no change occurs in the tonal melody of the stem. By contrast, if a H-toned object index is inserted (for example, the object index ba of class 2), the spreading of its H tone modifies the tone of the first two syllables of lexically toneless stems.

Some of the grammatical elements anteposed to the verb stem have tonal properties whose analysis is less obvious. For example, contrary to the future marker tlaa (which has no influence on the tonal melody of the verb stem at all), or to object markers (which have no influence on the tonal melody of lexically H-toned stems), the potential marker ka triggers the replacement of the first H tone of lexically H-toned stems by a L tone.

The tonal properties of verb finals

The tonal alternations affecting the last syllables of verb stems constitute the trickiest aspect of the tonal morphology of the languages of the Sotho-Tswana group, including Setswana. The tonal properties of the last syllable of verb stems vary from tense to tense in such a way that there is no way of accounting for them by means of general rules. As already illustrated in example (20) above, they are commonly involved in the distinction between conjoint and disjoint forms, but in a puzzling way, since the tonal properties that characterise the final syllable of the conjoint form in some tenses are those of the disjoint form in some other tenses. A detailed analysis of this aspect of the tonal morphology of Setswana can be found in Creissels (2017).

There is, in particular, an important distinction between two groups of tenses. In some tenses, when the verb occurs in non-final position in the sentence, the final syllable of the verb stem has a fixed tone regardless of the nature of the following word, whereas in other tenses, the final tone of the verb stem depends on the grammatical nature and tonal contour of the following word. For example, in the conjunct form of the present positive, the final tone of the verb form ke reka ‘I buy, I am buying’ changes if the object noun podi ‘goat’ is replaced by the demonstrative pronoun ele. By contrast, in the same conditions, no tonal change is observed in the corresponding negative form ga ke reke.

As already mentioned, the tonal transcriptions in this paper refer to the pronunciation of speakers of the Sengwaketse or Sengwato varieties of Setswana. Generally speaking, across the languages of the Sotho-Tswana group, there is little variation in the underlying tones of lexemes and grammatical words, or in the types of tonal processes that determine the tonal contour of words and sentences. However, there is important variation in the details of H tone spreading processes. For example, in Southern Sesotho, or in the Setlhaping or Serolong varieties of Setswana, H tone spreading is more limited than in the system described in the previous sections. This explains for example the difference in the tonal melody of the sentence Ke kotula mabele as pronounced by speakers of Southern Sesotho and by speakers of the Sengwaketse variety of Setswana. In Sesotho, the spreading of the H tone attached to the first syllable of the verb stem is limited to the following syllable, whereas in Sengwaketse, three successive syllables are involved in the spreading process.

• Sesotho:

  Example 25

  Ke kotula mabele.

  [kɩ̀-kótúlà màbɛ̀ːlɛ́]

  I am harvesting sorghum.

• Sengwaketse:
  Example 26

  Ke kotula mabele.

  [kɩ̀-kótúlá mábɛ̀ːlɛ]

  I am harvesting sorghum.

H tone spreading is particularly developed in the Sekgatla variety spoken in Mochudi. In the system presented in the previous sections, H tone spreading is strictly ‘local’ in the sense that it is always limited to a given number of syllables (between one and three, depending on the grammatical and phonological nature of the context in which H tone spreading occurs, but never more than three). By contrast, as illustrated in Table 3 and 4, the Sekgatla variety spoken in Mochudi has ‘unbounded’ H tone spreading in the sense that there is no limit to the length of the sequences of L-toned syllables that may be affected by the spreading of a H tone.

• Sengwaketse:
  Example 27

  Ke bofololela Mothibi dikgomo.

  [kɩ̀-bófʊ́lʊ́lɛ̀là́ mʊ̀tʰìbì dìqʰòːmʊ́]

  I am untying the cows for Mothibi.

• Sekgatla:
  Example 28

  Ke bofololela Mothibi dikgomo.

  [kɩ̀-bófʊ́lʊ́lɛ́lá mʊ́tʰíbí díqʰòːmʊ́]

  I am untying the cows for Mothibi.

What characterizes tone languages is the stability of the tonal melody of words with a given lexical and phonological content in a given phonological context. However, this stability is not contradictory with the possible existence of a complex system of rules that modify the tonal melody of words in given phonological and/or grammatical conditions, and a proper understanding of the tonal system of Setswana requires awareness of the articulation between the basic tonal melody of words and the two types of tone rules that determine or modify it:

• morphological rules determining the basic tonal melody of words depending on their lexical and grammatical content;
• rules operating at word boundaries or at the end of sentences (or at the end of each intonational phrase if the sentence is divided into two or more intonational phrases).

Crucially, a relatively straightforward account of the rules that determine or modify the tonal melody of words in Setswana requires acknowledging that the division of sentences into words as codified in the traditional orthography of Setswana is purely conventional, since many grammatical items traditionally written as separate words have a specific tonal behaviour that cannot be predicted by considering them as words.

The tonal morphology of verbs is particularly complex in Setswana, and the reader must be aware that, in this short sketch of Setswana tone, it was impossible to provide an exhaustive overview of this aspect of Tswana. However, I have tried to illustrate the various types of tonal processes accounting for the grammatically conditioned variation in the tone of verb forms.

Finally, in most of its aspects, the tonal system is essentially stable across Setswana dialects, with however an important exception: there is some dialectal variation in the details of the processes that can be analysed in terms of H tone spreading.

Creissels, D. (2017) ‘The conjoint/disjoint distinction in the tonal morphology of Tswana’. In Van der Wal, J. & Hyman, L. M. (eds.). The conjoint/disjoint alternation in Bantu. Boston: De Gruyter Mouton. pp. 200-238.