Navigation – Plan du site

AccueilNuméros9(1)Truncation, affection, and pain i...

Truncation, affection, and pain in Southern Bantu

Andre Batchelder-Schwab

Résumés

Cet article vise à rendre compte des propriétés phonologiques et discursives d’une troncation post-syntaxique utilisée en xhosa, tswana et venda pour rendre un caractère affectif avec les termes de parenté et la douleur avec les parties du corps. Des données élicitées en provenance de locuteurs natifs montrent que la troncation des syntagmes possessifs implique la suppression de la sonorité intervocalique à l’extrémité droite du syntagme, qui correspond généralement à l’indice d’accord de classe. Les conditions phonologiques et pragmatiques de la forme de possession réduite (FPR) sont similaires dans toutes les langues examinées, mais les possesseurs et possédés éligibles varient considérablement. La répartition actuelle dans la famille élargie des langues bantoues est inconnue, mais il existe des pistes prometteuses dans des langues comme le cuwabo et le kiswahili.

Haut de page

Entrées d’index

Haut de page

Texte intégral

1. Background

1What are we communicating when we swear or use slurs? When we use nicknames? When we use expressions like ouch! or oops!? These types of meaning have long been the subject of discussion among linguists and psychologists, but the general consensus has centered on these terms not carrying descriptive semantics, but rather emotional, expressive semantics (see Elffers 2014 for a summary since the 1600s). The term expressives is meant to unify linguistic phenomena that are largely dependent on emotional meaning. Work with aphasia patients, for example, shows that individuals who cannot formulate linguistic propositions can still curse well and curse often, leading to the conclusion that this process of producing emotion-based meaning must be occurring in a different part of the brain (Jay 1999).

  • 1 Ponsonnet’s work on Barunga Kriol (Australia) shows a stark contrast between a reduplication morph (...)

2Expressive and descriptive (truth-conditional) meanings have often been described as parallel systems (Kratzer 1999; Sawada 2013; Potts 2007). Many morphemes, such as English little, demonstrate simultaneous work on both fronts: little communicates smallness, and it can communicate endearment (Schneider 2003). Endearment has been linked to diminutives to the point that some have posited its universality (Strang 1968), despite many languages grammaticalizing affection through alternate strategies, like possession (Pukui & Elbert 1979, for Hawaiian). Other examples of simultaneous meaning are slurs like Kraut, a derogatory Anglophone term for Germans, which simultaneously conveys that a referent is German (descriptive) and that the speaker does not like Germans (expressive) (Gutzmann 2013). Despite occasional overlap with descriptive meaning, perhaps the most important aspect of expressives is that they are generally syntactically and semantically optional, meaning that sentences with and without them are both acceptable in traditional grammaticality judgments (Kratzer 1999; Fortin 2011: 68; Ponsonnet 2018),1 and do not add descriptive meaning but rather flesh out perspectives in the context of the utterance (Flier 1975: 2; Beard 1995).

3Returning to the example of Kraut, the argument here is that descriptive semantics and syntactic well-formedness are unaffected by the use of a slur instead of the unmarked alternative, German. These traits apply not only to aforementioned words like ouch and oops (Kratzer 1999), but also some pragmatic particles in German (Gutzmann & Turgay 2012), Korean honorifics which seem to tolerate unexpected syntactic mismatches (Song et al. 2019; Pak 2015), and binding problems with experiencer verbs (Pesetsky 1987).

  • 2 All uncited Hawaiian data throughout the paper was also elicited by the author from two high schoo (...)

4This paper intends to add to the current literature by delineating a truncation construction in three Southern Bantu languages: this truncation process holds an expressive meaning of affection towards a subclass of kinship terms (including some pets) and of pain in a subclass of body part terms. Unless otherwise cited, the data in this paper was elicited in sessions with one speaker each of isiXhosa, TshiVenḓa, Setswana, and a short session with one speaker of Kiswahili, all in Boston between 2020 and 2021.2 This data is supplemented by academic and media materials such as movies and books. The truncation of possessor phrases onto their possessee nouns will be labeled Reduced Possession Form (RPF), as it will be compared to languages which show similar patterns but whose usage may not be expressive and instead syntactic, like Cuwabo (Guérois 2015).

5The truncation constructions in possessive NPs are broadly similar in semantics, phonology, and structure across the examined Bantu languages, and are argued to underlyingly be the same process: a pragmatically licensed, post-syntactic truncation of an intervocalic sonorant at the right edge of the phrase. This sonorant is usually — but not always — the class agreement on the possessor phrase (for an overview of Bantu noun classes, see Katamba 2003). The licensability of the construction varies across the three languages, with isiXhosa (Section 2) having the most extensive list of both possessors and possessees in this truncation phenomenon, and thus the greatest analysability on a phonological and pragmatic level. TshiVenḓa (Section 3) has the narrowest set, with only two eligible possessees — mother and child — and one eligible possessor, the first person singular. Setswana (Section 4) will be shown to have a narrower set of eligible possessors and possessees, and will be compared to similar data from its sister language Northern Sotho (Mojapelo 2007). A discussion (Section 5) will show that the conditions for truncation are extremely similar across these Southern Bantu languages, both in phonology and semantics, and will examine similar data from Kiswahili and other Bantu languages in light of this. Finally, the conclusion (Section 6) will present a summary of current findings as well as diagnostics for finding similar constructions in other Bantu languages, which delineates how cognate constructions are predicted to surface, given the microvariation established within this paper.

2. IsiXhosa (S40)

6IsiXhosa is the southernmost Bantu language on the African continent, and forms the tail end of the Nguni dispersal southwards. The puzzle this paper will attempt to explain in isiXhosa and related languages is produced below. The truncation of the noun class agreement [w] in (2) is argued to be a spell-out of affectionate expressive semantics (Potts 2007).

(1) Ndithethe nomama wam.
ndi‑theth‑e na‑u‑mama u‑a‑m
1sg.sm‑speak‑recp with‑1‑mother 1‑poss‑1sg.poss
‘I spoke with my mother.’
(2) Ndithethe nomamam.
ndi‑theth‑e na‑u‑mama‑w‑a‑m
1sg.sm‑speak‑recp with‑1‑mother‑1‑poss1sg.poss
‘I spoke with my mother.’ (affectionate)

7The alternation is entirely syntactically and semantically optional, in almost the way that mother and mommy are interchangeable. According to the consultant, the reduced possession form (henceforth RPF) in (2) is much more affectionate, informal, close, and even child-like. This possession paradigm can be extended to other kinship terms, too, such as for ‘father’, ‘daughter’, ‘child’, ‘son’, ‘wife’, ‘husband’, ‘son-in-law’, ‘daughter-in-law’, ‘mother-in-law’, and ‘friend’, as the examples below show.

(3) utata wam utatam
u‑tata u‑a‑m
1‑father 1‑poss‑1sg.poss
‘my father’ ‘my dear father’
(4) intombi yam intombam
iN‑tombi i‑a‑m
9‑girl 9‑poss‑1sg.poss
‘my daughter’ ‘my dear daughter’
(5) umnt(w)ana wam umntanam
um‑ntwana u‑a‑m
1‑child 1‑poss‑1sg.poss
‘my child’ ‘my dear child’
(6) umkwenyana wam umkwenyanam
um‑kwenyana u‑a‑m
1‑son.in.law 1‑poss‑1sg.poss
‘my son-in-law’ ‘my dear son-in-law’
(7) unyana wam unyanam
u‑nyana u‑a‑m
1‑son 1‑poss‑1sg.poss
‘my son’ ‘my dear son’
(8) umhlobo wam umhlobam
um‑ɬobo u‑a‑m
1‑friend 1‑poss‑1sg.poss
‘my friend’ ‘my dear/best friend’
(9) umyeni wam umyenam
um‑yeni u‑a‑m
1‑husband 1‑poss‑1sg.poss
‘my husband’ ‘my dear husband’

8All of these examples again have expressive connotations of endearment, emotional closeness between the speaker and the relative, as well as having informal and almost child-like connotations. Note that this construction is not restricted to a single class, as the Class 9 example shows in (4). This form is DP-internal and not predicative in any sense. The independence of this construction from DP-external forces such as the verb or prepositions is demonstrated below, by placing both the unmarked possession form and the truncated possession form in varying thematic roles.

9Unergative

  • 3 This sentence is ambiguous between a present stative and a past tense, so (10) and (11) could also (...)
(10) Umama wam uncumile.3
u‑mama u‑a‑m u‑ncum‑ile
1‑mother 1‑poss‑1sg.poss 1sm‑smile‑recp
‘My mother smiled.’
(11) Umamam uncumile.
u‑mama‑u‑a‑m u‑ncum‑ile
1‑mother‑1‑poss‑1sg.poss 1sm‑smile‑recp
‘My dear mother smiled.’

10Unaccusative

(12) Umama wam uwile.
u‑mama u‑a‑m u‑w‑ile
1‑mother 1‑poss‑1sg.poss 1sm‑fall‑recp
‘My mother fell.’
(13) Umamam uwile.
u‑mama‑w‑a‑m u‑w‑ile
1‑mother‑1‑poss‑1sg.poss 1sm‑fall‑recp
‘My dear mother fell.’

11Theme

(14) Ndibone umama wam.
ndi‑bon‑e u‑mama u‑a‑m
1sg.sm‑see‑recp 1‑mother 1‑poss‑1sg.poss
‘I saw my mother.’
(15) Ndibone umamam.
ndi‑bon‑e u‑mama‑w‑a‑m
1sg.sm‑see‑recp 1‑mother‑1‑poss‑1sg.poss
‘I saw my dear mother.’

12Transitive agent

(16) Umama wam uthenge imoto.
u‑mama u‑a‑m u‑theng‑e i‑moto
1‑mother 1‑poss‑1sg.poss 1sm‑buy‑recp 9‑car
‘My mother bought a car.’
(17) Umamam uthenge imoto.
u‑mama‑w‑a‑m u‑theng‑e i‑moto
1‑mother‑1‑poss‑1sg.poss 1sm‑buy‑recp 9‑car
‘My dear mother bought a car.’

13Goal

(18) Ndinike umama wam imoto.
ndi‑nik‑e u‑mama u‑a‑m i‑moto
1sg.sm‑give‑recp 1‑mother 1‑poss‑1sg.poss 9‑car
‘I gave the car to my mother.’
(19) Ndinike umamam imoto.
ndi‑nik‑e u‑mama‑w‑a‑m i‑moto
1sg.sm‑give‑recp 1‑mother‑1‑poss‑1sg.poss 9‑car
‘I gave the car to my dear mother.’

14Beneficiary

(20) Ndiculele umama wam.
ndi‑cul‑el‑e u‑mama u‑a‑m
1sg.sm‑sing‑appl‑recp 1‑mother 1‑poss‑1sg.poss
‘I sang for my mother.’
(21) Ndiculele umamam.
ndi‑cul‑el‑e u‑mama‑w‑a‑m
1sg.sm‑sing‑appl‑recp 1‑mother‑1‑poss‑1sg.poss
‘I sang for my dear mother.’

15Object of preposition

(22) Ndihambe nomama wam.
ndi‑hamb‑e na‑u‑mama u‑a‑m
1sg.sm‑walk‑recp with‑1‑mother 1‑poss‑1sg.poss
‘I walked with my mother.’
(23) Ndihambe nomamam.
ndi‑hamb‑e na‑u‑mama‑w‑a‑m
1sg.sm‑walk‑recp with‑1‑mother‑1‑poss‑1sg.poss
‘I walked with my dear mother.’

16Possessor

(24) Inja kamama wam ityile.
iN‑ja Ø‑ka‑mama u‑a‑m i‑ty‑ile
9‑dog 9‑poss.1‑mother 1‑poss‑1sg.poss 9sm‑eat‑recp
‘My mother’s dog ate.’
(25) Inja kamamam ityile.
iN‑ja Ø‑ka‑mama‑w‑a‑m i‑ty‑ile
9‑dog 9‑poss.1‑mother‑1‑poss‑1sg.poss 9sm‑eat‑recp
‘My dear mother’s dog ate.’

17This list demonstrates that this alternation is entirely independent of theta role and the verb more broadly, and this holds true for the cognate constructions in related languages. Instead, given felicitous pragmatics, it is DP-internal constraints which largely govern applicability of RPF: the type of possessee, the type of possessor, and restrictions on phonological deletion.

18Regarding possessors, RPF is licit for all tested pronominal possessors in isiXhosa, as demonstrated below. Later data in Setswana and TshiVenḓa will establish a difference here in that their version of RPF is limited to 1sg possessors. This construction can apply to all the possessive pronouns of the language, including those for non-humans like ‘dogs’ (32).

(26) umama wam umamam
u‑mama u‑a‑m
1‑mother 1‑poss‑1sg.poss
‘my mother’ ‘my dear mother’
(27) umama wakho umamakho
u‑mama u‑a‑kho
1‑mother 1‑poss‑2sg.poss
‘your mother’ ‘your dear mother’
(28) umama wakhe umamakhe
u‑mama u‑a‑khe
1‑mother 1‑poss‑3sg.poss
‘his/her mother’ ‘his/her dear mother’
(29) umama wethu umamethu
u‑mama u‑a‑ithu
1‑mother 1‑poss‑1pl.poss
‘our mother’ ‘our dear mother’
(30) umama wenu umamenu
u‑mama u‑a‑inu
1‑mother 1‑poss‑2pl.poss
‘y’all’s mother’ ‘y’all’s dear mother’
(31) umama wabo umamabo
u‑mama u‑a‑bo
1‑mother 1‑poss‑3pl.poss
‘their mother’ ‘their dear mother’
(32) umama wazo umamazo
u‑mama u‑a‑zo
1‑mother 1‑poss‑10.poss
‘their mother’ (of dogs) ‘their dear mother’ (of dogs)
(33) umama waye umamaye
u‑mama u‑a‑ye
1‑mother 1‑poss‑9.poss
‘her mother’ (pronominalization of umama
wentombi
‘the mother of the girl’)
'her dear mother' (of
Class 9 subject)
  • 4 I switch away from using umama here because umakaSipho is licit, as is utakaSipho for ‘Sipho’s fat (...)

19Above, the experiencer of affection towards the possessee switches depending on the possessor. For example, in (31), the mother is unambiguously interpreted to be very close with ‘them’, and not the speaker. These examples make clear that it is possible for parties that are not the speaker or the listener to be involved in this construction for isiXhosa, which is not be the case in other languages. For isiXhosa, every noun capable of RPF can carry every type of possessor shown above. Nevertheless, isiXhosa does have restrictions on possessors for RPF, namely a categorical rejection of non-pronominal possessors, which is shared by the other Bantu languages for this construction.4

(34) umhlobo kaSipho *umhloboSipho//*umhlokaSipho
um‑hlobo Ø‑ka‑Sipho
1‑friend 1‑poss.1‑Sipho
‘Sipho’s friend’ *intended: ‘Sipho’s dear friend’
(35) umhlobo wenja *umhlobenja
um‑hlobo u‑a‑iN‑nja
1‑friend 1‑poss‑9‑dog
‘the dog’s friend’ *intended: ‘the dog’s dear friend’
(36) umhlobo wendoda *umhlobendoda
um‑hlobo u‑a‑iN‑doda
1‑friend 1‑poss‑9‑man
‘the man’s friend’ *intended: ‘the man’s dear friend’
(37) umama woluntu *umamuluntu//*umamoluntu//*umamawuntu
u‑mama u‑a‑ulu‑ntu
1‑mother 1‑poss‑9‑community
‘the community’s mother’ *intended: ‘the community’s dear mother’

20Having established that RPF in isiXhosa is only licensable for pronominal possessors, the next question is the extent of this paradigm in terms of eligible possessees. Recalling that RPF is licit for many kinship terms like ‘mother’, ‘father’, ‘daughter’, ‘child’, ‘son’, ‘son-in-law’, ‘friend’, and ‘husband’, not all kinship terms are eligible:

(38) umzala wam *umzalam
um‑zala u‑a‑m
1‑cousin 1‑poss‑1sg.poss
‘my cousin’ *intended: ‘my dear cousin’
(39) ubhuti wam *ubhutam
u‑bhuti u‑a‑m
1‑brother 1‑poss‑1sg.poss
‘my brother’ *intended: ‘my dear brother’
(40) umalume wam *umalumam
u‑malume u‑a‑m
1‑uncle 1‑poss‑1sg.poss
‘my uncle’ *intended: ‘my dear uncle’
(41) utat’omkhulu wam *utat’omkhulam/*utat’omkhulwam
u‑tat’omkhulu u‑a‑m
1‑grandfather 1‑poss‑1sg.poss
‘my grandfather’ *intended: ‘my dear grandfather’

21This cannot be attributed to pragmatics, as affection itself is possible towards any kinship member, and the phonological conditions in (38) for ‘cousin’ are virtually identical to the ones for ‘mother’. Thus, it is only a subset of kinship terms which are eligible for RPF on seemingly lexical grounds. This differs from plausible explanations for RPF restrictions on plural kinship terms on seemingly phonological grounds.

(42) oomama bam *oomamam//*oomabam
oo‑mama b‑a‑m
2b‑mother 2‑poss‑1sg.poss
‘my mothers’ *intended: ‘my dear mothers’
(43) abazali bam *abazalam//*abazabam
aba‑zali b‑a‑m
2a‑parents 2‑poss‑1sg.poss
‘my parents’ *intended: ‘my dear parents’
(44) abahlobo bam *abahlobam
aba‑hlobo b‑a‑m
2a‑friend 2‑poss‑1sg.poss
‘my friends’ *intended: ‘my dear friends’
(45) iintombi zam *iintombam//*iintozam
iziN‑ntombi z‑a‑m
10‑girl 10‑poss‑1sg.poss
‘my daughters’ *intended: ‘my dear daughters’
(46) amantombi ʔam amantombam
ama‑ntombi ʔ‑a‑m
6‑girl 6‑poss‑1sg.poss
‘my daughters’ ‘my dear daughters’

22The crucial piece of data here is (45) and (46), where two pluralization strategies exist in modern isiXhosa: (45) which is standard pluralization in Class 10, and (46) which is non-standard but common pluralization in Class 6. This variation allows a clear demonstration of how phonological restrictions work in the language, where an easily deletable consonant like the glottal stop [ʔ] is eligible for RPF, while other consonants like [z] are ineligible. Kinship terms are highly restricted in noun class, so further proof of this phonological restriction will be shown in body parts which appear in more of the noun classes.

23Initial attempts to use RPF for body parts resulted in failure when attempting to put a body part into an affection construction. However, the following movie scene revealed how RPF is actually used for body parts in isiXhosa.

24Context: A man has just been hit by a car, and his leg broke.

(47) umlenze wam!
um‑lenze u‑a‑m
3‑leg 3‑poss‑1sg.poss
‘my leg!’
(48) umlenzam!
um‑lenze‑w‑a‑m
3‑leg‑3‑poss‑1sg.poss
‘ouch, my leg!’ (Qubeka 2018)

25The scene above takes place in the 2018 movie Sew the Winter to My Skin when the South African police are firing into a crowd of protesters. One man is run over by a vehicle; his leg snaps grotesquely and he calls out with (48). The example in (47) is given as the unmarked version of this phrase, for reference. Upon review with my consultant, this type of truncation with body parts is indeed productive for scenarios with pain, surprise, and anger. This construction can take some other body parts, as shown below.

(49) umnwe wam umnwam!
um‑nwe u‑a‑m
3‑finger 3‑poss‑1sg.poss
‘my finger’ ‘Ow, my finger!’
(50) amashiye am amashiyam!
ama‑shiye ʔ‑a‑m
6‑eyebrow 6‑poss‑1sg.poss
‘my eyebrows’ ‘my eyebrows!’
(51) umzimba wam umzimbam!
um‑zimba u‑a‑m
3‑body 3‑poss‑1sg.poss
‘my body’ ‘ow, my body!’
  • 5 One might be puzzled as to why the output is not [umlomwam] given the underlying final vowel of /u (...)
(52) umlomo wam umlomam!5
um‑lomo u‑a‑m
3‑mouth 3‑poss‑1sg.poss
‘my mouth’ ‘ow, my mouth!’
(53) intloko yam intlokwam! (*intlokam)
iN‑ɬoko i‑a‑m
9‑neck 9‑poss‑1sg.poss
‘my neck’ ‘ow, my neck!’
(54) impumlo yam! impumlwam!
iN‑phumlo i‑a‑m
9‑nose 9‑poss‑1sg.poss
‘my nose’ ‘ow, my nose!’
(55) ingqiniba yam ingqinibam!
iN‑ngqiniba i‑a‑m
9‑elbow 9‑poss‑1sg.poss
‘my elbow’ ‘ow, my elbow!’

26It is licit to exclaim (50) if one’s eyebrows are burning off, or if they were largely removed by the aesthetician. A similar usage exists in English, one can scream my leg! when hit by a vehicle, or shout my eyebrows! if they were accidentally threaded off in a salon. These are not expressions of affection towards the body parts, yet this usage, some form of ouch!, has been described as lying within expressive semantics as part of a natural class with affection (Kratzer 1999). It is also the case that body parts and kinship form a crosslinguistic natural class as inalienable nouns (Heine 1997), a discussion which we will return to in Section 5. However, many body parts cannot undergo RPF on seemingly phonological grounds, and any body part whose class concord is not a glide or glottal is completely ineligible for this structure.

(56) izandla zam *izandlam
izi‑andla zi‑a‑m
8‑hand 8‑poss‑1sg.poss
‘my hands’ *intended: ‘ow, my hands!’
(57) ubuso bam *ubusam//*ubuswam
ubu‑so bu‑a‑m
14‑face 14‑poss‑1sg.poss
‘my face’ *intended: ‘ow, my face!’
(58) unyawo lwam *unyawam//*unyalwam
ulu‑nyawo lu‑a‑m
11‑foot 11‑poss‑1sg.poss
‘my foot’ *intended: ‘ow, my foot!’
(59) isisu sam *isisam//*isiswam
isi‑su si‑a‑m
7‑stomach 7‑poss‑1sg.poss
‘my stomach’ *intended: ‘ow, my stomach!’

27This fact is more clearly demonstrated by the following pairs of body parts which match in root. However, the different agreement markers in the plural change the acceptability for RPF, where Class 5 items with a lateral l‑ prefix on the possession phrase are categorically barred from RPF, while their plural counterparts in Class 6 with a glottal ʔ‑ prefix are allowed. Conversely, Class 9 singular body parts with a glide y‑ prefix allow RPF, while their plural counterparts in Class 10 with a fricative z‑ prefix are categorically barred from RPF.

(60) iqatha lam *iqatham//*iqalam
ili‑qatha li‑a‑m
5‑ankle 5‑poss‑1sg.poss
‘my ankle’ *intended: ‘ow, my ankle!’
(61) amaqatha ʔam amaqatham!
ama‑qatha ʔ‑a‑m
6‑ankle 6‑poss‑1sg.poss
‘my ankle’ ‘ow, my ankles!’
(62) indlebe yam indlebam!
iN‑dlebe i‑a‑m
9‑ear 9‑poss‑1sg.poss
‘my ear’ ‘ow, my ear!’
(63) iindlebe zam *iindlebam!//*iindlezam!
iiN‑dlebe z‑a‑m
10‑ear 10‑poss‑1sg.poss
‘my ears’ *intended: ‘ow, my ears!’

28Finally, there are some noun classes like 3‑4 in which both agreement prefixes are glides, and so RPF is eligible for both singular and plural body parts, as shown below.

(64) umnwe wam umnwam!
um‑nwe u‑a‑m
3‑finger 3‑poss‑1sg.poss
‘my finger’ ‘ow, my finger!’
(65) iminwe yam iminwam!
imi‑nwe i‑a‑m
4‑finger 4‑poss‑1sg.poss
‘my fingers’ ‘ow, my fingers!’

29The RPF restrictions for body parts make clear that there is a strong phonological component at play: the agreement prefix must be either a glide y or w or must be a glottal stop ʔ in order to be eligible for deletion. Below Table 1 summarizes the data so far by displaying a sample word possessed by the first person singular with the class concord prefix in bold; licit reduced forms are in italics, while illicit reduced forms are starred and non-italic. All unshaded forms are the licit realization of the regular, unmarked possession structure in the language.

Table 1 ― Truncated possession (reduced possession form) in nouns classes in isiXhosa

Class Full RPF Full RPF
Singular Plural
1‑2 ‘friend’ umhlobo wam umhlobam abahlobo bam *abahlobam
3‑4 ‘finger’ umnwe wam umnwam iminwe yam iminwam
5‑6 ‘eye’ iliso lam *ilisam amehlo ʔam amehlam
7‑8 ‘hand’ isandla sam *isandlam izandla zam *izandlam
9‑10 ‘ear’ indlebe yam indlebam iindlebe zam *iindlebam
11‑10 ‘foot’ unyawo lwam *unyawam iinyawo zam *iinyawam
14 ‘face’ ubuso bam *ubusam --
15 ‘food’ ukutya kwam *ukutyam --

Italics: truncated possession is possible; regular: truncated possession is not possible

30A key case demonstrates truncation where the noun class is not easily deletable, and RPF seems to skip to the next intervocalic consonant to the right:

  • 6 Although this is isiXhosa data, an isiZulu film entitled Nabantwa bam (literally: With my children (...)
(66) abantwana bam abant(w)abam6 (*abantwanam)
aba‑ntwana b‑a‑m
2‑child 2‑poss‑1sg.poss
‘my children’ ‘my dear children’
  • 7 Because alienability seems to play a part in RPF licensability, my initial analysis was that no cl (...)

31If this is the same type of RPF — and it seems to be on pragmatic grounds — this effect on the root itself confirms RPF in isiXhosa as a deletion process instead of a lack of agreement in the syntax. This was the only strong example in the data where the deletion was not of a noun-class prefix on the possessor phrase. Building from this, then, if the deletion usually affects the class concord consonants, then the agreement relation must have happened prior to triggering the RPF deletion process. We can deduce that RPF must be post-syntactic, happening after noun class agreement is complete between the root and the possession phrase.7

32Further proof that RPF must be a post-syntactic process comes from the examples with ‘home’ below concerning and adds weight to a theory that alienability is playing a role in RPF licensing:

(67) indlu yam *indl(w)am
in‑ɮu i‑a‑m
9‑house 9‑poss‑1sg.poss
‘my house’ *intended: ‘my dear house’
(68) endl(w)ini yam endlinam
e‑in‑ɮu‑ini i‑a‑m
loc‑9‑house‑loc 9‑poss‑1sg.poss
‘at my home’ ‘at my beloved home’
  • 8 Further reading: Taraldsen (2020).

33Note that RPF is only eligible on the locative-circumfixed8 ‘home’, and not on the unmarked ‘house’; this is an idiosyncrasy with no explanation yet. What is clear from the truncated form in (68) is that the possession is clearly outside the locative suffix ‑ini, indicating that the syntactic locativization must have happened prior to RPF, adding further complication to claiming early-attachment inalienable possessors in explaining Bantu RPF.

34It must be noted that Example (68) is extremely unusual in allowing RPF. The examples so far have solidified a phonological requirement for the RPF, but this is clearly not the only factor, as the previous ineligible kinship terms made clear. The attempted RPF for ‘cousin’ is reproduced in (69), as well as several other phonologically acceptable environments. Alongside the phonological considerations for RPF, the vast majority of nouns are ineligible for RPF in any capacity when they are not either kinship, body parts, or the home, specifically examples like (70)‑(72), which have favorable phonological contexts for RPF deletion.

(69) umzala wam *umzalam
um‑zala u‑a‑m
1‑cousin 1‑poss‑1sg.poss
‘my cousin’ *intended: ‘my dear cousin’
(70) ukrebe wam ?*ukrebam
u‑krebe u‑a‑m
1‑shark 1‑poss‑1sg.poss
‘My shark’ *intended: ‘my dear shark’
(71) umfula wam *umfulam
um‑fula u‑a‑m
3‑stream 3‑poss‑1sg.poss
‘my stream’ *intended: ‘my dear stream’
(72) incwadi yam *incwadam
iN‑ncwadi i‑a‑m
9‑book 9‑poss‑1sg.poss
‘my book’ *intended: ‘my dear book’

35Inanimate objects are almost categorically excluded from reduction, but it is worth noting that my classmate researched sharks and loved them very much, and so by the end of the semester, (70) playfully became licit, although this cannot be expected to be a typical grammatical judgment outside of this context. However, this does reveal a narrow productivity for animals which could be considered either pets or friends.

(73) ikati yam ikatam
i‑kati i‑a‑m
9‑cat 9‑poss‑1sg.poss
‘my cat’ ‘my dear cat’
(74) intaka yam intakam
iN‑taka i‑a‑m
9‑bird 9‑poss‑1sg.poss
‘my bird’ ‘my dear bird’
(75) inja yam ?injam
iN‑nja i‑a‑m
9‑dog 9‑poss‑1sg.poss
‘my dog’ ‘my dear dog’
(76) inyoka yam ?*inyokam
iN‑nyoka i‑a‑m
9‑snake 9‑poss‑1sg.poss
‘my snake’ *intended: ‘my dear snake’
(77) umkhaza wam *umkhazam
um‑khaza u‑a‑m
3‑tick 3‑poss‑1sg.poss
‘my tick’ *intended: ‘my dear tick’

36The analysis so far predicts that all of the above RPF should be theoretically allowed, at least on phonological grounds. However, animals which are not pets are illicit for the structure, specifically (76) and (77). When presented with a scenario in which the author has four pet snakes, (76) became marginal, but the consultant said that the practice of keeping snakes as pets is culturally infelicitous among amaXhosa and (76) would be strange on those grounds. Cultural infelicity was the same reason behind the marginality of (75), where the consultant said that dogs were not traditionally given affection or considered pets, but an example with proper context allows (75), as will be shown in (82). RPF for (77) was completely unacceptable regardless of contextualizing the tick as a pet, which demonstrates a gradient acceptability scale for animals. Despite not being prototypical kinship terms, certain pets seem eligible for RPF, ostensibly elevating them to kinship as has been attested in other languages (Tsunoda 1996: 576). This evokes the prior acceptability for ‘my dear friend’ (8), umhlobam, which might not be kinship in the strictest sense.

37Beyond the lexicon and phonology, there are pragmatic considerations for RPF which seem to influence acceptability. A prototypical example of RPF in discourse is presented below.

38Context: Speaker’s mother passed away a few years ago

(78) Ndikhumbula umamam.
ndi‑khumbul‑a u‑mama‑w‑a‑m
1sg.sm‑miss‑fv 1‑mother‑1‑poss‑1sg.poss
‘I miss my dear mother (my mom).’
(79) Ndikhumbula umama wam.
ndi‑khumbul‑a u‑mama u‑a‑m
1sg.sm‑miss‑fv 1‑mother 1‑poss‑1sg.poss
‘I miss my mother.’

39In this context, the speaker is explicitly expressing affection towards her mother, and the possession structure in (78) is an overt marker of this qualification of closeness. Although not syntactically incorrect, the context can pragmatically rule out the more neutral sentence (79).

  • 9 Note that if we were to mark the object (mother) on the verb, it would spell out as 3sg despite se (...)

40Context: A mother teasing her grown son after his wedding9

(80) Ngubani oza kukhathalela umamakho ngoku?
ngubani a‑u‑za ku‑khathal‑el‑a u‑mama‑u‑a‑kho ngoku
who rel‑3sm‑fut inf‑care‑appl‑fv 1‑mother‑1‑poss‑2sg.poss now
‘Who will take care of your mommy now?’
(81) (#)Ngubani oza kukhathalela umama wakho ngoku?
ngubani a‑u‑za ku‑khathal‑el‑a u‑mama‑u‑a‑kho ngoku
who rel‑3sm‑fut inf‑care‑appl‑fv 1‑mother1‑poss‑2sg.poss now
‘Who will take care of your mother now?’

41In a similar way, (80) and (81) are very close in meaning and are syntactically interchangeable. Yet, (81) does not explicitly encode a conversational move being made by the mother here, one of affirmation of her closeness with her son. The affective form here implies teasing, which is a well-known form of reaffirming closeness (Blythe 2012; Haugh & Pillet-Shore 2018). The closeness is emphasized by RPF, and in fact its absence in (81) could be strange and could be a misinterpreted as distance or real concern about the matter. That is to say, using the construction in (80) explicitly rules out readings of pragmatic distance between the mother and her addressee.

42Context: A woman treats her ugly dog like a baby, and we’re being mean about it.

(82) Injakhe imbi!
in‑ja‑i‑a‑khe im‑bi
9‑dog‑9‑poss‑3sg.poss cop.9‑ugly
‘Her little doggie is ugly!’
(83) Inja yakhe imbi!
in‑ja i‑a‑khe im‑bi
9‑dog 9‑poss‑3sg.poss cop.9‑ugly
‘Her dog is ugly!’
  • 10 This is similar to the use of endearment possessors in Hawaiian for mocking, specifically the seco (...)

43Returning to the discussion on pets, we can recall that ‘dog’ was only marginally acceptable in (75), yet becomes licit in the context of (82). This statement crucially relies on making explicit the emotional closeness between her and her dog, only to turn around to mock this fact. According to the speaker’s intuition, (82) places the owner and the dog into the roles of parent and child. Again like before, it is not semantically or syntactically wrong to pick one structure over the other, but there are solid pragmatic implications that the shorter affectionate structure in (83) makes overt — specifically teasing.10

44As before, if pets count as kin, then isiXhosa displays a fascinatingly prototypical alienability effect for RPF, where body parts, some kin, and the home are eligible to the exclusion of the rest of the nouns of the language, when the phonological and pragmatic licensing conditions are met. This pattern is much clearer for isiXhosa than the following languages, given that RPF is much less productive in TshiVenḓa and Setswana.

3. Affection in TshiVenḓa (S21E)

45TshiVenḓa, alongside a few close language varieties like Xironga, form a divergent branch of Southern Bantu languages in the extreme northeast of South Africa in the Limpopo state, as well as across the border in Zimbabwe. It is classified as its own node S20 in the Southern Bantu language group (Maho 2009), which makes it a cousin to isiXhosa. The data analyzed in this section was collected from a speaker of the Nzhelele TshiVhase dialect, which is tentatively classified as S21E. Recalling the truncation pattern delineated for isiXhosa, the alternation below should look familiar:

(84) wana wanga ṋwananga
nw‑ana u‑a‑nga
1‑child 1‑poss‑1sg.poss
‘my child’ ‘my dear child’
  • 11 The TshiVenḓa speaker could easily envision children using this construction productively. This i (...)

46In spoken language, this alternation is easy to miss, and could preliminarily be explained as a purely phonological process which deletes intervocalic w. As with isiXhosa, the forms above are interchangeable syntactically and semantically, there is no context which causes one to be exclusively grammatical or felicitous and not the other. It also seems to be unique to this lexical item, none of the other words in this noun class were able to shorten like this, in the judgment of this speaker. Example (84) has very specific connotations, however: it is almost child-like,11 it is informal, and it communicates love for the child, almost like a nickname. The RPF can be used by an adult addressing the child, or calling the child to come over, in introducing the child with affection to someone else, and to assert authority and ownership of the child. This last point can and has been used in a scenario where a man is trying to claim his biological child, although he is not involved in the child’s upbringing. In choosing this form, the father is explained to be affirming his love and closeness to the child by choosing RPF over the unmarked possession form.

47Unlike for isiXhosa, only the first person singular possessor is eligible for RPF.

(85) ṋwana washu *ṋwanashu
nw‑ana u‑a‑shu
1‑child 1‑poss‑1pl.poss
‘our child’ *intended: ‘our dear child’
  • 12 The possessive pronoun ‑vho is homophonous for both 2sg.poss and 3pl.poss. Neither accept RPF.
(86) wana wahvo12 *ṋwanavho
nw‑ana u‑a‑vho
1‑child 1‑poss‑2sg.poss/3pl.poss
‘your/their child’ *intended: ‘your/their dear
child’
(87) ṋwana waṋu *ṋwanaṋu
nw‑ana u‑a‑nu
1‑child 1‑poss‑2pl.poss
‘y’all’s child’ *intended: ‘y’all’s dear child’
(88) ṋwana wawe *ṋwanawe
nw‑ana u‑a‑we
1‑child 1‑poss‑3sg.poss
‘his/her child’ *intended: ‘his/her dear child’

48Thus, the truncated construction in TshiVenḓa is restricted to the first person singular possessor. With the parameter of person restriction set for possessors in TshiVenḓa, we turn to investigate the extent of the paradigm. The only other possible truncated form found in the language is the term for ‘mother’ in Class 1a (as in Mulaudzi 1996: 31).

(89) mme ʔanga mmeanga
m‑me ʔ‑a‑nga
1‑mother 1a‑poss‑1pl.poss
‘my mother’ ‘my dear mother’
(90) baba ʔanga *babanga
Ø‑baba ʔ‑a‑nga
1a‑father 1a‑poss‑1pl.poss
my father’ *intended: ‘my dear father’
(91) maḽumi ʔanga *maḽumanga
Ø‑maḽumi ʔ‑a‑nga
1a‑uncle 1a‑poss‑1pl.poss
‘my uncle’ *intended: ‘my dear uncle’

49Thus, the entirety of the truncation phenomenon in TshiVenḓa centers on the words for ‘mother’ and ‘child’ being possessed by the first person singular. Neither body parts nor the term for home can be truncated, and thus this constitutes the most narrow data set examined by this paper.

4. Soweto Setswana (S31)

50Setswana forms part of a larger dialect continuum with Northern Sotho and counts well over 8 million speakers (Eberhard et al. 2022). Keeping in mind the prior truncation pattern found in TshiVenḓa and isiXhosa, the alternation in possession of kinship terms below again looks familiar.

(92) ngwana wame ngwaname
ngw‑ana u‑a‑me
1‑child 1‑poss‑1sg.poss
‘my child’ ‘my dear child’
(93) mma wame mmame
m‑ma u‑a‑me
1‑mother 1‑poss‑1sg.poss
‘my mother’ ‘my dear mother’
(94) tsala wame tsalame
Ø‑tsala u‑a‑me
1a‑friend 1‑poss‑1sg.poss
‘my friend’ ‘my dear/best friend’
(95) motswala wame motswalame
mo‑tswala u‑a‑me
1‑cousin 1‑poss‑1sg.poss
‘my cousin’ ‘my dear cousin’
(96) koko wame kokwame
Ø‑koko u‑a‑me
1a‑grandma 1‑poss‑1sg.poss
‘my grandmother’ ‘my dear grandmother’
(97) monna wame monname
mo‑nna u‑a‑me
1‑man 1‑poss‑1sg.poss
‘my husband’ ‘my dear husband’

51The full form with two separate words is the unmarked possessive form which is the most standard way of speaking and writing; the truncation again gives connotations similar to a nickname, and is judged to be child-like, affectionate, and informal. It was not obvious at first whether to include koko (96) ‘grandmother’ given the ambiguous [w] that could originate either as a phonological effect from the last letter of the lexical item for grandmother or be from the agreement on the possessive phrase. The judgment of the consultant clarifies that there is a definite meaning change between the two variants in (96), which is exactly comparable with the other pairs. The data above constitutes the entire set of kinship terms eligible for RPF, at least for this speaker.

52Note that there are several asymmetries for eligible possessees in Setswana, specifically that ‘husband’ is licit for this form but not ‘wife’, that ‘grandmother’ is licit but not ‘grandfather’, and that ‘mother’ is licit but not ‘father’. Idiosyncrasy in alienability classes is attested crosslinguistically: the Hawaiian inalienable o-class contains words such as ‘canoe’ and ‘chair’ while excluding ‘child’ (Elbert 1970: 52); Mokilese treats ‘vehicle’ as morphologically inalienable while ‘car’ is alienable (Harrison 1976: 124, 128); Northern Pomo treats most kinship terms inalienably, but excludes ‘relative’ (O’Connor 1992: 257). Given such data from languages with explicit alienability splits in the morphology, an alienability system in Setswana which includes ‘husband’ but not ‘wife’ is at least plausible.

53For sake of clarity, (98) shows an ungrammatical attempt at placing an alienable object into this construction, despite the favorable phonological environment. This is the result for every other noun in the language, except ‘stomach’, which will be covered in (103).

(98) nama yame *namame
N‑nama y‑a‑me
9‑meat 9‑poss‑1sg.poss
‘my meat’ (i.e. of a chicken) *intended: ‘my dear meat’

54Constructions not using the first person singular possessor were judged as ungrammatical, as shown below, and extend to all other possessors. Recall that this is the same restriction as was shown for TshiVenḓa, and is unlike isiXhosa.

(99) monna wago *monnago
mo‑nna u‑a‑go
1‑man 1‑poss‑2sg.poss
‘your husband’ intended: ‘your dear husband’
(100) monna wage *monnage
mo‑nna u‑a‑ge
1‑man 1‑poss‑3sg.poss
‘her husband’ intended: ‘her dear husband’
  • 13 My consultant specifically judged this one as ungrammatical, which could come down to either a dia (...)

55This list of terms in Setswana is largely corroborated by work on Northern Sotho (Mojapelo 2007: 123), a closely related variety where the identical corresponding words for mma ‘mother’, motswala ‘cousin’, and moko ‘grandmother’ from above are also being licit in a truncated possession structure, as well as papa ‘father’, mmatswale ‘mother-in-law’, ratswale ‘father-in-law’, rakgolo ‘grandfather’13 and motlogolo ‘nephew/niece’. Mojapelo goes a step further than my consultant, however, and gives the example below.

(101) Ke rata [mmago].
ke rata m‑ma‑wa‑ga‑go
1sg love 1‑mother‑1‑poss.1‑1sg.poss
‘I love [your mother].’ (glosses mine)
  • 14 The Northern Sotho judgments are drawn from Mojapelo, who is herself a native speaker, as well as (...)

56A footnote explicitly explains that “[mmago] is the contracted form of [mma [wa gago]]” (Mojapelo 2007: 150; bold in the original changed to italics). Note that this means that Northern Sotho speakers clearly allow for possessors outside of the first-person singular, specifically the second person.14 The analysis of these kinship terms is very brief, and Mojapelo does not explain whether there is a meaning difference between the longer and shorter versions, but uses this evidence to strengthen claims of inalienable syntax in the language. This leaves the extent of eligible possessors in Northern Sotho affection truncation an open question, but at minimum includes the speaker and addressee. The brevity of Mojapelo’s treatment of this specific construction in Northern Sotho leaves open questions about whether the truncated form (101) has endearment in its meaning.

57On the other hand, Setswana clearly uses RPF for the same set of expressive meaning as isiXhosa and TshiVenḓa. A very close variety like Northern Sotho could be predicted not to stray too far in this respect, but this remains conjecture for now. The Setswana consultant gave the following contexts for the form ngwaname ‘my dear child’:

(102) Contexts for ngwaname:
a. My child, I need you to learn to pray before I die.
b. My child loves me! (He surprised me with a visit or a gift)
c. My child said something hurtful to me.
d. My child is sick and I am scared.
e. My child graduated! (Proud; maybe it has been a struggle)

58RPF is not licit without a strong emotional component in Setswana, matching previous patterns. Using the truncated form gives the intention behind it “something extra”, and is a statement of love from the speaker’s standpoint. Thus, a small subset of kinship terms in Setswana expand the possessive truncation phenomenon found in TshiVenḓa beyond the concept of child, but maintains person and number restrictions for possessors which are stricter than isiXhosa.

59The final piece of relevant data for Setswana is presented below, which shows the truncated possession applied to a body part. This was the only body part which was found to license the construction in the language. The context for (103) would be to have a stomach ache, and the consulted speaker notes that pregnant women are known to say this when the baby kicks. The alignment with the isiXhosa usage for pain is noteworthy and adds further proof of some alienability effect for RPF.

(103) mpa yame! mpame!
N‑pa i‑a‑me
9‑stomach 9‑poss‑1sg.poss
‘my stomach’ ‘ow, my stomach!’

60My consultant did not accept any other nouns as eligible possessees for RPF, including house or pets, but she suggested that other speakers might. Again like for isiXhosa, the major limitation of this study is that the data is gathered from a single speaker per language. However, the RPF pattern emerging across these languages behaves similarly in phonology, pragmatics, and lexical specification.

5. Discussion

61The crosslinguistic analysis of the affectionate truncation phenomenon is further spelled out in this section (Table 2). On a phonological level, isiXhosa, Setswana and TshiVenḓa largely pattern together in mostly deleting glides and glottal stops. Included here is the isiXhosa data point in which RPF unusually skips a syllable and deletes an alveolar nasal for ‘my children’.

Table 2 — Truncatable consonants in different languages

Deleted
consonant
w y ʔ n
isiXhosa umntwana wam
‘my child’
→ umntwanam
intombi yam
‘my daughter’
→ intombam
amashiya ʔam
‘my eyebrows’
→ amashiyam
abantwana bam
‘my children’
→ abantabam
Setswana ngwana wame
‘my child’
→ ngwaname
mpa yame
‘my stomach’
→ mpame
TshiVenḓa wana wanga
‘my child’
→ wananga
mme ʔanga
‘my mother’
→ mmeanga

62This class of consonants /w j n ʔ/ could form a natural class of sounds to the exclusion of other sounds like /l b s k z/. The Sound Pattern of English treats glottal consonants as sonorants based on lack of constriction (Chomsky & Halle 1968), giving some precedent for treating a glottal stop as a sonorant. The roadblock might be the [l], which is typically considered sonorant, but some have argued that lateral liquids can be nonsonorant depending on the language (Yip 2005). The prime evidence for this explanation would be Setswana, in which /l/ is allophonically realized as [d] before high vowels (Gouskova et al. 2011). However, categorization of exclusively [w j n ʔ] by the term “sonorant” is simply a matter of convenience, and the rule could simply be further specified against laterals. Thus, a possible truncation rule could be formulated as a single application of the rule below.

(104) [TRUNC]: V[+sonorant]V → Ø/_V(- sonorant)(V)C(V)#

63This analysis does not yet contend with how exactly this truncation is initiated by syntax or pragmatics, given that meaning change must reside outside phonology in many frameworks. Rather, this paper serves to bring to light the existence of this phenomenon, its contexts of use, its phonological characteristics, and its post-syntactic operation. However, its restricted distribution only within the realm of inalienable possession is noteworthy and points towards the necessity of accounting for syntax in a full analysis.

64It is important to note that the RPF pragmatics across the languages are very similar: all three consultants independently mentioned that there is something “child-like” about the construction. Affectives characterized as a feature of child-speak is a recurring feature of this type of expressive construction crosslinguistically, perhaps best attested in diminutives, to the point that diminutives are proposed to universally derive from the word for ‘child’ (Jurafsky 1996). Russian poetry can feature ik for child imagery (Kalina-Levine 1981: 37); diminutives are a salient component of politeness in child-directed speech in Macedonian (Spasovski 2012); Colloquial Jordanian Arabic diminutives convey pragmatics that fundamentally revolve around children (Badarneh 2010: 156). This perception is easily substantiated in English by listing terms with affective suffix /‑i/ (horsey, birdey, kitty, piggy, sheepie), which has been long noted as a feature of English child speech (Svaib 1993; Schneider 2003). However, many Bantu languages already have a diminutive (isiXhosa, Setswana, TshiVenḓa: ‑ana), which has been convincingly shown to etymologically stem from the word for ‘child’ and have true size semantics as well as similar expressive affection semantics (Jurafsky 1996; Fortin 2011). In contrast, RPF has no size semantics and cannot be etymologically derived from ‘child’, yet speakers derive a child-like association despite this. Also noteworthy is that the narrowest RPF paradigm examined in this paper, TshiVenḓa, critically only includes the words for ‘child’ and ‘mother’. Hawaiian use of affective possessors also displays the same judgments of child-speech from speakers, despite also having no size semantics and no obvious derivation etymologically from ‘child’. This association can hardly be coincidental.

  • 15 Schneider (2003) also mentions handy, with over 140,000 hits, but whose homophony with the more com (...)

65The isiXhosa and Setswana use of RPF for pain when directed to body parts is a link that lends further evidence to treating these expressive meanings as part of some natural class (Kratzer 1999). Further evidence includes pain class words (i.e. hurt) significantly associating on a scale with pleasure class (i.e. loveable) in computational work with corpora (Grefenstette et al. 2006). Hawaiian again has a very similar pattern where expressive possession of body parts is used for ouch! contexts, further validating the crosslinguistic tendency. In comparison with English /‑i/ again, we find tummy as an expressively marked version of ‘stomach’ (Schneider 2003: 91), revealing that English, too, has a similar construction which can be used in pain constructions and has child-register connotations. This term tummy is far more common in a corpus search (17,405 hits) than Schneider’s other attested terms like footie (2,114), leggy (3,909) and toothie (11) (NOW corpus: Davies 2013).15 It is likely then that English, when using /‑i/ for pain constructions, has a similar distribution to Setswana’s single RPF body term ‘stomach’, revealing that such narrow usage in the lexicon is probably not a rare phenomenon. Crosslinguistic analysis reveals this systematic pattern which might otherwise be explained as a singular idiosyncrasy of a language’s lexicon.

66RPF possessors in isiXhosa, Setswana, and TshiVenḓa are restricted to pronouns. IsiXhosa accepts all pragmatically viable pronouns, including non-human ones like for umamazo ‘their dear mother (of dogs)’. RPF licensing in TshiVenḓa and Setswana, on the other hand, is restricted to first person singular possessors according to my data. Northern Sotho, a close variety to Setswana, clearly accepts the second person singular possession in truncated possession, but no meaning difference has been established (Mojapelo 2007).

67The final point of comparison is the lexical restrictions on licensing. Some languages like Wailaki (Begay 2018: 226), Hawaiian, and English seem to have no real lexical restrictions on possessees of affection constructions, given sufficient pragmatic licensing in context. English speakers frequently coin new /‑i/ terms like stimmy-check for ‘stimulus check’, or Dunky Donuts for ‘Dunkin Donuts’. Conversely, this does not seem to be the case for the Bantu language data examined in this paper so far, whose RPF licensing is word-specified and not similarly productive in discourse. IsiXhosa did demonstrate marginal productivity for pets, but this stands in stark contrast to the overwhelming productivity of affective constructions regardless of possessee alienability in Hawaiian and English. Every Bantu RPF possessee was at least arguably inalienable in this paper, which would predict a similar pattern elsewhere, yet Kiswahili data almost immediately breaks this mold.

68Kiswahili has been documented to have syntactically optional constructions called “contractions” that look similar to the RPF of Southern Bantu (Mpemba 2015). Mpemba describes them as functionally serving, “as a brevity, time-saving, and speech connectivity device” (2015: 1). Some are reproduced below from the paper.

(105) a. babao (instead of baba yao ‘their father’)
b. mwanetu (instead of mwana wetu ‘our son/child’)
c. jinale (instead of jina lake ‘his/her/its name’)
d. kakayo (instead of kaka yako ‘your (sg.) brother’)
e. chumbache (instead of chumba chake ‘his room’)
f. nyumbenu (instead of nyumba yenu ‘your (pl.) house’)
g. kambangu (instead of kamba yangu ‘my rope’)
h. wemao (instead of wema wako ‘your kindness’)
i. kitabucho (instead of kitabu chako ‘your book’)
j. majinangu (instead of majina yangu ‘my names’)
k. ugonjwangu (instead of ugonjwa wangu ‘my disease’)

69The vast majority of possessed nouns are not eligible for contraction, according to Mpemba. Interestingly, Kiswahili also includes a wider variety of deletable consonants and semantic categories, specifically a word like ‘rope’, which is not crosslinguistically predicted to pattern inalienably. There does not seem to be a restriction on personhood in pronouns, as the examples above include 1st, 2nd and 3rd person pronouns in both the singular and plural, though Mpemba does show that 3sg is much more contractable than the others, in what might simply be a separate non-expressive paradigm (Mpemba 2015: 10). Mpemba makes no note of semantic differences between contracted and uncontracted forms, and attributes the phenomenon as resulting from frequency, where high-frequency terms tend to be contracted more often.

70A frequency-based approach is not necessarily incompatible with one of alienability, given that possessed inalienable nouns tend to be high-frequency. In fact, the frequency of possession of body parts and kinship terms has been argued as a primary driver of alienability splits in grammars crosslinguistically (Haspelmath 2008: 3). However, a consultant for Kiswahili confirms that babangu ‘my dear father’ is a more affectionate form of baba yangu ‘my father’, and likewise for kakayo ‘your dear brother’ and mwanao ‘your dear child’. Testing the rest of the paradigm in Kiswahili for meaning differences in the contractions constitutes further research, but the attested meaning difference for the contraction in kinship terms is non-trivial and shows that this might constitute a cognate structure to the previously discussed Southern Bantu languages.

71Finally, there are attested Bantu Reduced Possession Forms (RPF) which phonologically seem similar to what has been laid out in this paper, with contracted possessor phrases onto their possessee nouns, but with major functional differences which would necessitate a fundamentally different syntactic explanation. One example is Mpyemo (A86c) in which possessor phrases for kinship and the home are obligatorily contracted, but only for the 1sg possessor (Paul Murrell pers. comm. 2021). Without meaning difference derived through an alternation, this would likely not serve an expressive function: RPF in Mpyemo would simply be morphologically specified on the first person possessive pronoun.

72Similarly, Cuwabo (P34) systematically cliticises all pronominal kinship possessors onto the possessee noun instead of placing them in a separate agreement phrase, which does not seem to constitute a pragmatically governed set of alternatives (Guérois 2015: 217). Additionally, Cuwabo has a vowel length reduction in possessive phrases during fast speech, which could be called RPF under its current definition (Guérois 2015: 214), illustrating yet another possible type of conditioning environment that exists across the Bantu family. For example, speech rate might constitute at least part of the explanatory power behind the Kiswahili paradigm given its description as optional phonological contraction to save time (Mpemba 2015: 4). Again, these are speculations that future work on both these and related languages might clarify.

6. Conclusion

  • 16 Data for contraction in Kiswahili was gathered from the data tables in Mpemba (2015: Appendix). No (...)

73This paper has documented a truncation process in possessed nouns in Southern Bantu languages which serves to denote expressive meaning — affection towards kin, or pain for body parts. Eligibility for the truncation seems somehow tied to alienability, given that only kin, body parts, and the home have been found to take this structure. The list of nouns eligible for truncation is language-specific, ranging from a couple (TshiVenḓa) to dozens (isiXhosa). There is also variation in eligible possessors, where isiXhosa licenses all persons and numbers, while Setswana and TshiVenḓa only license 1sg possessors (and Northern Sotho may also license 2sg). The deletable consonants are argued to form a natural class of sonorants, and may occur one or two syllables from the right edge of the word. A summary of the points can be found in Table 316, in which a few crosslinguistic tendencies will be outlined in hopes of helping future research in this area.

Table 3 — Compared features in languages discussed in this paper

Family Language Possessee Possessor Expressive Structure
Bantu (RPF) isiXhosa some kin
body parts
house.loc
all pronouns yes contraction:
[n w j ʔ]
Setswana some kin
stomach
1sg yes contraction:
[w j]
N. Sotho some kin
?body parts
1sg, 2sg ? contraction:
[w j χ]
TshiVenḓa mother
child
1sg yes contraction:
[w ʔ]
Kiswahili some kin
few objects
all pronouns yes, some contraction:
[w j l ʧ z]
Cuwabo some kin all pronouns no -
Mpyemo kin
house
1sg no -
Austronesian Hawaiian open class 1sg, 2sg
(archaic)
yes noun class
Dene Wailaki open class 1sg yes pronoun
Indo-European English open class open class yes suffix /‑i/
Macedonian open class open class yes diminutive
Russian open class open class yes diminutive

74First, glides and glottal stops are deleted more often than stops and fricatives, which shows a trend along sonority hierarchy. Second, RPF and other morphological expressives in possession are most common in the first person singular possessor, evidenced by Setswana, TshiVenḓa, modern Hawaiian and Wailaki, followed by the second person in Northern Sotho and archaic Hawaiian, then followed by the full pronominal paradigm in isiXhosa and Kiswahili. This could be outlined as a [1>2>3] crosslinguistic person hierarchy (Rosen 1990: 704 for Southern Tiwa; Hockett 1939 for Potawatomi), where we predict that if a language allows for a 3rd person RPF, then it also allows for 2nd and 1st person. In the context of Bantu, the possessee class most often eligible for RPF is close kinship, followed by body parts, the home, and then by alienable objects. This predicts that RPF is most likely to appear in other languages on inalienable possessees, though this probably does not need to be the case given Kiswahili. Lastly, RPF might not always serve the same role in the grammar: a relatively unified use for expressives has been covered in detail by this paper, but Cuwabo and Mpyemo hint at variation in what RPF might accomplish in other languages.

75Although this paper has not outlined a detailed syntactic analysis for the process which seems simultaneously linked to expressive semantics, phonology and alienability syntax, a descriptive account of RPF is intended to bring attention to this overlooked issue. It is currently unknown how many other languages have RPF, though it has appeared in every Bantu language that I have examined. In the same vein, it is unclear whether these paradigms in languages like isiXhosa and Kiswahili constitute cognate syntax inherited from their common ancestor, whether they constitute multiple independent developments, or whether endearment syntax is laterally transferable. Transfer from neighboring languages would be especially plausible for TshiVenḓa due to narrow RPF usage and alternative method (noun class) for similar expressive meanings (Mulaudzi 2000), which the other languages lack. Further factors could be TshiVenḓa’s relatively small speaker population and the extreme multilingualism of the South African sprachbund. The limitation to these hypotheses is lack of data, specifically lack of access to more than a single speaker each of these languages. However, planned follow-up studies in South Africa with a broader number of speakers will hopefully clarify some of these questions.

Abbreviations

Ø null morpheme
1 Class 1 marking (idem for 2‑15)
10.poss Class 10 possessive pronoun
1pl first person plural
1sg first person singular
2a first version of noun Class 2, aba‑
2b second version of noun Class 2, oo‑
2pl second person plural
2sg second person singular
3pl third person plural
3sg third person singular
9.poss Class 9 possessive pronoun
9sm Class 9 subject marking
appl applicative
cop.1 copula for Class 1
cop.9 copula for Class 9
fut future
fv final vowel (used in isiXhosa present tense)
inf infinitive
loc locative
poss possession morpheme
poss.1 possession morpheme for Class 1 possessors
recp recent past
rel relativizer
RPF reduced possession form
sm subject marking
wh question word
Haut de page

Bibliographie

Babou, Cheikh Anta & Michele Loporcaro. 2016. Noun classes and grammatical gender in Wolof. Journal of African Languages and Linguistics 37(1). 1‑57.

Badarneh, Muhammad A. 2010. The pragmatics of diminutives in colloquial Jordanian Arabic. Journal of Pragmatics 42(1). 153‑167.

Beard, Robert. 1995. Lexeme-morpheme base morphology. Albany, NY: SUNY Press.

Begay, Kayla Rae. 2018. Wailaki grammar. Berkeley, CA: University of California (PhD dissertation) (accessed 26 May 2023).

Blythe, Joe. 2012. From passing-gesture to ‘true’ romance: Kin-based teasing in Murriny Patha conversation. Journal of Pragmatics 44(4). 508‑528.

Chomsky, Noam & Morris Halle. 1968. The sound pattern of English. New York, NY: Harper & Row.

Collins, Chris & Paul M. Postal. 2012. Imposters: A study of pronominal agreement. Cambridge, MA: MIT Press Scholarship Online.

Davies, Mark. 2013. Corpus of news on the Web (NOW): 3+ billion words from 20 countries, updated every day.

Eberhard, David M., Gary F. Simons & Charles D. Fennig (eds.). 2022. Ethnologue: Languages of the world. 25th ed. Dallas, TX: SIL International.

Elbert, Samuel H. 1970. Spoken Hawaiian. University of Hawaii Press.

Elffers, Els. 2014. Exclamatives: A grammatical category? History and Philosophy of the Language Sciences.

Flier, Michael S. 1975. Remarks on Russian verbal prefixation. The Slavic and East European Journal 19(2). 218.

Fornander, Abraham. 1917. Hawaiian antiquities and folk-lore (Bernice P. Bishop Museum Memoires 4‑5). Honolulu, HI: Bishop Museum Press.

Fortin, Antonio. 2011. The morphology and semantics of expressive affixes. Oxford: University of Oxford (PhD dissertation).

Gouskova, Maria, Elizabeth Zsiga & One Tlale Boyer. 2011. Grounded constraints and the consonants of Setswana. Lingua 121(15). 2120‑2152.

Grefenstette, Gregory, Yan Qu, David A. Evans & James G. Shanahan. 2006. Validating the coverage of lexical resources for affect analysis and automatically classifying new words along semantic axes. In James G. Shanahan, Yan Qu & Janyce Wiebe (eds.), Computing attitude and affect in text: Theory and applications, vol. 20, 93‑107. Berlin: Springer.

Guérois, Rozenn. 2015. A grammar of Cuwabo (Mozambique, Bantu P34). Lyon: Université Lumière Lyon 2 (PhD dissertation).

Gutzmann, Daniel. 2013. Expressives and beyond: An introduction to varieties of use-conditional meaning. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives: Explorations in use-conditional meaning, 1‑58. Leiden: Brill.

Gutzmann, Daniel & Katharina Turgay. 2012. Expressive intensifiers in German: Syntax-semantics mismatches. In Christopher Piñón (ed.), Empirical issues in syntax and semantics, vol. 9, 1‑18.

Harrison, Sheldon P. 1976. Mokilese reference grammar. Honolulu, HI: University Press of Hawaii.

Haspelmath, Martin. 2008. Alienable vs. inalienable possessive constructions. Course “Syntactic universals and usage frequency” at the Leipzig Spring School on Linguistic Diversity. Leipzig, ms.

Haugh, Michael & Danielle Pillet-Shore. 2018. Getting to know you: Teasing as an invitation to intimacy in initial interactions. Discourse Studies 20(2). 246‑269.

Heine, Bernd. 1997. Possession: Cognitive forces, sources, and grammaticalization. Cambridge: Cambridge University Press.

Hockett, Charles. 1939. Potawatomi syntax. Language 15(4). 235.

Jay, Timothy. 1999. Why we curse: A neuro-psycho-social theory of speech. Amsterdam: John Benjamins.

Jurafsky, Daniel. 1996. Universal tendencies in the semantics of the diminutive. Language 72(3). 533.

Kalina-Levine, Vera. 1981. Through the eyes of the child: The artistic vision of Elena Guro. The Slavic and East European Journal 25(2). 30.

Katamba, Francis. 2003. Bantu nominal morphology. In Gérard Philippson & Derek Nurse (eds.), The Bantu languages, 103‑120. London: Routledge.

Kratzer, Angelika. 1999. Beyond ouch and oops. How descriptive and expressive meaning interact: A comment on David Kaplan’s paper. Paper presented at the Cornell Conference on Theories of Context Dependency.

Maho, Jouni Filip (comp.). 2009. NUGL online: The online version of the new updated Guthrie list, a referential classification of the Bantu languages (accessed 26 May 2023).

Mojapelo, Mampaka Lydia. 2007. Definiteness in Northern Sotho. Stellenbosch: University of Stellenbosch (PhD dissertation).

Mpemba, Titus. 2015. The formation and syntax of contractions in Kiswahili with special emphasis on noun-possessive combination. Journal of Linguistics and Language in Education 9(2). 1‑31.

Mulaudzi, Phalandwa Abraham. 1996. A descriptive analysis of the morphology of the Tshiguvhu dialect of Venda. Pretoria: University of South Africa (M.A. thesis) (accessed 26 May 2023).

Mulaudzi, Phalandwa Abraham. 2000. The ‘two way’ formation: the diminutive nouns in the Tshiguvhu dialect of Venda. South African Journal of African Languages 20(2). 159‑164.

Myler, Neil. 2016. Building and interpreting possession sentences. Cambridge, MA: MIT Press Scholarship Online.

Nxumalo, Kulile. 2004. Nabantwa Bam. Real Stories from a Free South Africa 5. Film.

O’Connor, Mary Catherine. 1992. Topics in Northern Pomo grammar. New York, NY: Garland.

Pak, Miok D. 2015. Towards understanding the syntactic representation of honorifics in Korean. Korean Linguistics 17(2). 132‑166.

Pesetsky, David. 1987. Binding problems with experiencer verbs. Linguistic Inquiry 18(1). 126‑140.

Ponsonnet, Maïa. 2018. Expressive values of reduplication in Barunga Kriol (northern Australia). Studies in Language 42(1). 226‑255.

Potts, Christopher. 2007. The expressive dimension. Theoretical Linguistics 33(2). 165‑198.

Pukui, Mary Kawena & Samuel H. Elbert. 1979. Hawaiian grammar. Honolulu, HI: The University Press of Hawaii.

Qubeka, Jamil X. T. 2018. Sew the winter to my skin. Film.

Rosen, Carol. 1990. Rethinking Southern Tiwa: The geometry of a triple-agreement language. Language 66(4). 669‑713.

Sawada, Osamu. 2013. The meanings of diminutive shifts in Japanese. In Stefan Keine & Shayne Sloggett (eds.), Proceedings of North East Linguistic Society 42, 163‑176.

Schneider, Klaus P. 2003. Diminutives in English. Tübingen: Max Niemeyer.

Song, Sanghoun, Jae-Woong Choe & Eunjeong Oh. 2019. An empirical study of honorific mismatches in Korean. Language Sciences 75. 47‑71.

Spasovski, Lupco. 2012. Morphology and pragmatics of the diminutive: Evidence from Macedonian. Tempe, AZ: Arizona State University (PhD dissertation).

Strang, Barbara. 1968. Modern English structure. 2nd ed. London: Arnold.

Svaib, Trisha. 1993. Are doggies really little? Evidence from children’s use of diminuitives [sic]. In Eve V. Clark (ed.), The proceedings of the 24th Annual Child Language Research Forum, 149‑161. Stanford, CA: The Center for the Study of Language and Information Publications.

Taraldsen, Knut Tarald. 2020. The syntactic structure of the Xhosa e‑ N ‑ini locatives. Quaderni di Linguistica e Studi Orientali (QULSO) 6. 221‑237.

Tsunoda, Tasaku. 1996. The possession cline in Japanese and other languages. In Hilary Chappell & William McGregor (eds.), The grammar of inalienability. Berlin: De Gruyter Mouton.

Yip, Moira. 2005. Variability in feature affiliations through violable constraints: The case of [lateral]. In Marc van Oostendorp & Jeroen van de Weijer (eds.), The internal organization of phonological segments, 63‑92. Berlin: De Gruyter Mouton.

Haut de page

Notes

1 Ponsonnet’s work on Barunga Kriol (Australia) shows a stark contrast between a reduplication morpheme that is non-optional (telicity marker) and a separate reduplication morpheme that only provides expressive meaning of endearment, and which is completely optional syntactically.

2 All uncited Hawaiian data throughout the paper was also elicited by the author from two high school teachers of Hawaiian, one of whom is a native speaker. All elicitation was held online due to the Covid-19 pandemic.

3 This sentence is ambiguous between a present stative and a past tense, so (10) and (11) could also be interpreted as ‘My mother is smiling’. For the sake of clarity, we will take the recent past tense reading.

4 I switch away from using umama here because umakaSipho is licit, as is utakaSipho for ‘Sipho’s father’; these look like umkaSipho, which would mean ‘Sipho’s wife’. However, ma‑ operates like a prefix in the language, like in the name of the musician Madosini, a woman who is of the Dosini clan. The two terms for ‘mother’ and ‘father’ are unusual in this regard, and the pragmatics of umakaSipho and utakaSipho do not seem expressive like other RPF forms examined for isiXhosa. It cannot be ruled out that this is a form of RPF, but uncertainty leads me to exclude umakaSipho and utakaSipho from the analysis.

5 One might be puzzled as to why the output is not [umlomwam] given the underlying final vowel of /umlomo/. This is due to a language-wide dispreference for adjacent labials, resulting in deletion of the glide.

6 Although this is isiXhosa data, an isiZulu film entitled Nabantwa bam (literally: With my children) (Nxumalo 2004) shows that a cognate construction exists in isiZulu, and clearly implies affection according to an isiZulu speaker, who describes the comitative preposition na‑ as also adding to the affection. The film is about strong emotional bonds in a family.

7 Because alienability seems to play a part in RPF licensability, my initial analysis was that no class agreement happens due to lower possessor attachment for inalienable possession (Myler 2016). However, the data in (66) caused that analysis to collapse. It would also be possible to dismiss this example as a lexification, which is unappealingly close to ignoring inconvenient data.

8 Further reading: Taraldsen (2020).

9 Note that if we were to mark the object (mother) on the verb, it would spell out as 3sg despite semantically being 1sg. The mismatch is likely the same phenomenon as the “imposters” of Collins & Postal (2012).

10 This is similar to the use of endearment possessors in Hawaiian for mocking, specifically the second-person ‘your dear _’, a pronoun no longer in use in modern Hawaiian. I reproduce the line of a chant from Fornander (1917‑1918: 57), which is analyzed as a father mocking his son-in-law to his daughter (Pukui & Elbert 1979: 117): ʻEʻoe kū ka hauna lāʻau a kāua i kāne, he kōlea kāne, he wāwae liʻiliʻi, he ʻūlili kāne, he holoholo kahakai ‘Our war strokes are not suitable for your (fine) husband, your (fine) husband is a plover, a small legged [one]; your (fine) husband is a tattler, running along the beach.’

11 The TshiVenḓa speaker could easily envision children using this construction productively. This is further discussed in Section 5.

12 The possessive pronoun ‑vho is homophonous for both 2sg.poss and 3pl.poss. Neither accept RPF.

13 My consultant specifically judged this one as ungrammatical, which could come down to either a dialectal or individual difference. This is an area for future work.

14 The Northern Sotho judgments are drawn from Mojapelo, who is herself a native speaker, as well as a number of consultants who are also native speakers (2007: 8).

15 Schneider (2003) also mentions handy, with over 140,000 hits, but whose homophony with the more common adjective is misleading in an unrefined corpus search.

16 Data for contraction in Kiswahili was gathered from the data tables in Mpemba (2015: Appendix). Note on Austronesian noun class: Hawaiian 1sg possessors kaʻu ‘my (alienable)’, koʻu ‘my (inalienable)’, and kuʻu ‘my (affectionate)’ break down into D-NC-Possessor, where the middle vowel agrees with noun class. Noun classes which serve expressive function are not in the scope of this paper, but have been described in Shona and Fula (Fortin 2011), TshiVenḓa (Mulaudzi 2000) and Wolof (Babou & Loporcaro 2016: 18).

Haut de page

Pour citer cet article

Référence électronique

Andre Batchelder-Schwab, « Truncation, affection, and pain in Southern Bantu »Linguistique et langues africaines [En ligne], 9(1) | 2023, mis en ligne le 30 juin 2023, consulté le 19 avril 2026. URL : http://journals.openedition.org/lla/6651 ; DOI : https://doi.org/10.4000/lla.6651

Haut de page

Auteur

Andre Batchelder-Schwab

Boston University

Haut de page

Droits d’auteur

CC-BY-SA-4.0

Le texte seul est utilisable sous licence CC BY-SA 4.0. Les autres éléments (illustrations, fichiers annexes importés) sont susceptibles d’être soumis à des autorisations d’usage spécifiques.

Haut de page
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search