The semantics of three valency-changing verbal derivational suffixes in Nandi (a Nilo-Saharan language spoken in East Africa) are investigated. Two cases of polysemy and one of homonymy are found and related to general case theory (DeLancey 2000, Jakobson 1936, 1990) as well as to the general theory of concept formation (Rosch and Mervis 1996).
Cet article présente une étude de la sémantique des suffixes dérivationnels verbaux qui changent de valence en nandi (une langue nilo-saharienne de l’Afrique de l’Est). Deux cas de polysémie et un cas d’homonymie ont été trouvés et considérés par rapport à la théorie générale des cas (DeLancey 2000, Jakobson 1936, 1990) et à la théorie générale de la formation des concepts (Rosch et Mervis 1996).
Corps de l’article
1. Introduction 
The Nandi language is spoken in the highlands of western Kenya and belongs to the Nilotic branch of the Nilo-Saharan language family. Its basic word order is VSO with VOS a freely occurring variant. There are two tonally-distinguished cases, nominative and oblique. The former will be identified in transcription with the label (nom) . All nouns not marked with nom are oblique-case forms.
Nouns are inflected for number. Verbs are inflected for subject (generally only non-3rd person), object, negation, tense (3 degrees of past tense), mood and aspect (perfective/imperfective). Verbal derivation is a typical feature of the languages of East Africa. Although perhaps best known in the Bantu languages, it is also found in many, if not most, of the Nilotic languages. Nandi (Southern Nilotic) is not an exception, and there exist derived verbs of Ventive, Itive, Dative, Instrumental, Stative, Associative, Reciprocal, Causative, Detransitive, etc. types. As the names imply, these processes are associated with semantic (e.g. directional) and grammatical (valency-changing) meanings. What the names do not imply, however, is the multifunctionality and polysemy/homonymy of verbs derived with a given element. For example, the Instrumental is often found with a valency-changing role which cannot possibly be instrumental, and the Itive sometimes has a comitative meaning . I have never forgotten the first time I experienced the shock that comes when a totally unexpected meaning appears out of nowhere. In this case the morpheme /-ta/ had been entirely associated in my experience with the sense ‘Motion away’ (see 4.1 below) when I encountered the following example of a totally different sense:
In this paper I shall look in detail at the semantics associated with verbs formed by means of three derivational suffixes. All three can result in the addition of a single argument to the argument structure of a verb. As labels I shall use initially the traditional terms used in Creider and Creider 1989, Rottland 1982, Tucker 1964, and Tucker and Bryan 1966. Sometimes as well it will be convenient to refer to the morphemes by their shapes. It should be understood that these are fairly abstract representations. An additional goal will be to investigate the implications that this semantics has for the classification of semantic participant types or roles. The three suffixes are:
2. Dative /-ci/
2.1 Dative proper
The first class of meanings to consider may be termed instances of what in traditional grammar is termed the ‘dative proper’. “The dative proper denotes that to or for which something is or is done.” (Smyth 1956: 338).
Benefactives are very typical instances:
Malefactives also occur:
2.2 Locative dative
Note that in the previous examples, the ‘indirect object’ or ‘benefactee’ was human. In the following examples, the added valent is semantically a location:
Here is an example where the same verb illustrates both dative proper and locative meanings:
3. Instrumental /-e:/
3.1 Instrumental proper
The conventional term for derivation with this affix is highly appropriate in a large number of cases:
3.2 Non-instrumental meanings
In addition to the foregoing clearly instrumental examples, however, there are a very large number of non-instrumental meanings associated with this suffix. In all cases these can be translated with an English preposition.
3.2.4 ‘for (the amount of)’
3.2.6 ‘of, with respect to’
The following example gives both senses, one a non-instrumental sense of the preposition ‘with’:
This extremely wide range of prepositional senses is perplexing both because of its breadth and because it seems to include the locative prepositional senses found with the Dative suffix and even (25) the dative proper sense. The first phenomenon has a natural and interesting explanation in the diachronic development of the suffix. The /-e:/ suffix is etymologically related to an independent preposition, /e:ng/. Since this is almost the only preposition in the language (/àk/ ‘with, and’ and /a:p/ ‘of’ are the only others), it of necessity expresses a large number of senses which are distinguished in English:
The preposition may also be used in an instrumental sense, although this usage seems to be more marked:
4. Itive /-ta/
4.1 Itive proper
Verbs derived with this suffix may have a ‘thither’ directional sense. Its former name was ‘Motion away’. Note that the directional use does not change the valency of a verb.
In many cases, the Itive contrasts with the Ventive, a suffix with a ‘hither’ directional component which was formerly known as ‘Motion towards’ .
4.2 Comitative Itive
The Itive suffix also occurs as a valency-changing suffix where the added argument is glossed as ‘with’. It is unlike a typical comitative because in all cases that I know of the added argument is inanimate. Although certain uses might appear to be instrumental, this is due to the polysemy of English ‘with’. None of the following examples have an instrumental participant except those derived with /-e:/.
Compare the instrumental:
5. Analysis and Summary
The vast majority of verbs in Nandi are monovalent or bivalent. Monovalent verbs include both the ‘unaccusative type’ and the ‘unergative type’ although these distinctions are not formally marked and both verb types have nominative subjects.
Bivalent verbs include both result type transitives and non-result transitives (Fillmore 1971, DeLancey 2000) but the distinction is again not marked formally.
Most interestingly, trivalent verbs (ditransitives) are almost non-existent. One example has a theme and a locative role (in addition to an agentive role):
In some cases diathetic alternations (Levin 1993) allow the same verb to express more than two roles (although not in the same expression of course):
But in most cases, resort must be had to derivational suffixation to obtain a trivalent verb. In the classic trivalent verb for English ‘give’, it is interesting that as expected /-ci/ is used (/ki:ka:-cì/), and there is no underived verb form.
Since the majority of monovalent and bivalent verbs are associated with agent and theme roles, the contrast between underived and derived verbs is also very much a contrast between these two roles as opposed to all other roles. This matter is discussed below in 5.1 and 5.4.
Derivation with /-ci/ has the effect of adding a participant to the set of participants associated with a verb. Although this additional participant must therefore be added in any classification of participant types, what is most salient is that this participant does not in fact participate in the event or action denoted by the verb. Both ‘dative proper’ and ‘locative’ share this semantic characteristic of being the non-participant end-point or terminus of the event or action . If the added participant is human, then a dative proper meaning is found, one which shades very quickly over into the reason for the action or event. With inanimate objects or locations, the added entity is simply the locus for the action/event. It does not seem appropriate to analyze this division in terms of a single prototype (basically one or the other of the two) because there is no evidence that one is more basic than the other. Rather, this is a case of equal polysemy – two different but related senses.
Since a location can be spoken of as in some sense receiving an action, one could perhaps use the term ‘recipient’ as a label for the abstract sense, but I feel that this usage gives an undue priority to the ‘dative proper’ sense. Similarly the term ‘goal’ unduly emphasizes the teleological sense, which I think is secondary. Hence I shall use the term terminal as a label for the abstract sense of what this derivation adds to the semantics of verbs.
In work on basic semantic roles done by Larsen 1979, verbs are seen as having one, two or three valencies which are related to the semantic roles of agent, patient and beneficiary/locative. Recently this same claim has been made again (independently) in the context of universals of case (DeLancey 2000). DeLancey claims that there are three universal semantic case roles: Agent, Theme (= Patient) and Location. These roles are defined in relation to basic predicate types and hence are equivalent to thematic roles (DeLancey 2000: 3). There is strong support for the first two roles from Nandi, but very little for the Location role. This, however, is in terms of basic predicate types. When we turn to derived verbs, we note immediately that verbs derived with /-ci/ add precisely beneficiary/locative roles (or Location in DeLancey’s terminology).
If we analyze the examples in 3.2 in terms of semantic roles, there is an astounding variety:
These participants are neither agents nor patients, and there is no clear general term for this collection of participant-types in the literature, a fact bemoaned by Talmy 1985 specifically for Fillmore’s 1977 location, source, goal and path, and also illustrated by thematic hierarchies such as that of Jackendoff 1972: 43 (50) and Grimshaw 1990: 7-8 (51):
In an effort to rectify this situation, Talmy 1985: 139 proposed the term Ground for these classes of participants with motion verbs; however, Ground does not include instrument.
Croft 1991 used the term ‘intermediate thematic roles’ for a collection of roles including comitative, instrument, manner, means, benefactive/malefactive. Croft’s term ‘intermediate’ is not particularly felicitous for those roles which are endpoints, and ‘oblique’ is probably best left as a syntactic term, so we must invent a term of our own for non-agents (and non-experiencers) and non-patients. There is a subtle difference between participants added via /-e:/ and those added via /-ci/. The former are typically not end-points but rather seem to be involved in the further (with respect to the sense of the underived verb) specification of the action or event. Thus they have an adverbial sense, and therefore I suggest the label peripheral . Note that with source roles, although the result of the action is that the patient is removed from the source participant, the function of the derivation is to add this removed participant to the argument structure of the verb and the action/event frame. In all other cases, the participant is added to and remains in the context of the action.
In view of the diversity of roles, we should consider whether the overall semantic category associated with /-e:/ is best understood as an instance of definition by family resemblance (AB, BC, CD, DE) where “each item has at least one, and probably several, elements in common with one or more items, but no, or few, elements are common to all items” (Rosch and Mervis 1996: 443). It is clear that certain roles have more in common with each other than they do with other roles, e.g. means, instrument and amount; or source, goal, location, beneficiary, path (all sharing a locative element), and this suggests a hierarchical sense tree with senses related by hyponymy (isa). This can be thought of as a refinement of the notion of a flat ‘family’ of senses.
It is very difficult to see a connection between the itive and the comitative senses, and I presume that the suffix represents a conflation of two separate suffixes (as is the case, for example, with the dative case of Ancient Greek, which conflates formerly separate dative, instrumental and locative cases). Strong support for this separation comes from the fact that itive /-ta/ never changes valency  whereas comitative /-ta/ always does. There is then homonymy within the range of meanings associated with the suffix.
The fact that the comitative participant is never human provides a clue to the analysis of the semantics of the derived verb. The added element is never agentive; it is also not peripheral or terminal. Rather it is a co-theme.
With Itive /-ta/, we have a clear case of homonymy in that the directional use of this suffix is unrelated to its role in providing a new, non-animate participant. In addition, we have a type of participant which does not appear in any of the literature I have referred to. The role, however, is very clear. Defined negatively, it is non-agentive, non-terminal and non-peripheral. Defined positively, it is a co-theme.
5.4 Summary and discussion
Here is a brief summary of the analyses which have been proposed: for Dative /-ci/ it was found that there were two distinct types of participant: human recipients and inanimate locations. Both were action or event end-points, and the term terminal was proposed for the semantic role which the added participant played.
These data, rather than providing evidence for a single dimension of participant-type (as with Jackendoff and Grimshaw), seem to suggest a more complex hierarchy. Participants fall into two categories, core and non-core. Core participants (agents and patients) are typically directly represented in the semantics of simplex (non-derived) verbs. Non-core participants of considerable semantic subtlety (co-themes, terminals, and peripherals) are created and accessed via the lexical process of verbal derivation.
The three suffixes that we have studied share a small semantic space, that of an additional participant which is not central or required to be associated with a verb. It is not surprising that the meanings of the suffixes sometimes appear to overlap, and it should also not surprise us if other languages do not express the semantics of non-central participants in identical ways. That is, we are dealing with a matter of language-specific subtlety, although because of the small semantic space involved we should expect similar semantic notions to be grammaticized in other languages. As an example, we may consider Swahili, an unrelated Bantu language. In Swahili, as in Nandi, verbs may be derived via suffixation, and one of the suffixes found is termed the ‘prepositional’ or the ‘applied’. Among the semantic roles associated with it are benefactive, malefactive, itive and instrumental roles (Ashton 1947: 217-220). These roles are parceled out among three different suffixes in Nandi and none of these suffixes are associated solely with any single one of the roles.
Finally, we may note the kinds of semantic relations (Lyons 1977) involved. We have three different kinds of meaning relationships: senses which are equal subsenses of a more inclusive sense (recipient, location as subsenses of terminal), senses which are related in terms of hyponymy (means as a hyponym of instrument as a hyponym of peripheral), and senses which are unrelated (homonymy with the directional and co-theme senses of /-ta/). It is striking that the semantics of these grammatical elements is so similar to that of lexical items.
In his famous Kasuslehre article, Jakobson 1936, 1990 argues that although grammatical cases may not have a single primary or basic meaning («Grundbedeutung», «signification fondamentale»), they do have a general meaning («Gesamtbedeutung», «signification générale»), which is independent of context, and special or particular meanings («Sonderbedeutungen»), which are context-dependent. The analysis I have given for /-ci/ fits this scheme perfectly with the general meaning being ‘terminal’ and the two specific meanings of ‘location’ and ‘dative proper’ associated with the animacy status of the referent which has been added to the context. Although more complex because of the number of subsenses and their hierarchical relationships, the same can be said of the analysis of /-e:/. However, Jakobson’s framework does not fit that of the two /-ta/’s (nor the dative case of Ancient Greek as previously noted).
- I would like to thank Jane Tapsubei Creider for help with the data dealt with in this paper and Iver Larsen, Doris Payne, Mechthild Reh, and three RQL reviewers for suggestions for its revision.
- The orthography is phonemic except that an [ATR] distinction among vowels is not indicated. /ng/ and /ny/ are velar and palatal nasals respectively. /n/ before /k/ is velar. Tones are /á/ high, /â/ high-falling, /à/ low-falling, /a/ (unmarked) low. See Creider and Creider 1989 for additional details.
- Terms with an initial upper-case letter, e.g. “Instrumental”, are labels for derived verb types. Terms entirely in lower-case, e.g. “instrumental”, are concepts.
To the best of my knowledge this is the first time that such an investigation has been undertaken for any Nilotic language. I would like to emphasize the productivity of these derivational processes: all meanings illustrated in this paper occur with many other verbal roots (see Creider and Creider 2001 for additional examples). At the same time, although it is too far afield from the main topic of this article to discuss in detail, I wish to claim that these derived verbs are separate, although related, lexical items with respect to corresponding underived verbs. In support of that claim, here are some examples of very idiomatic usages of the suffixes. If the suffixes were available only for combinatorial use, then these frozen meanings would presumably not develop (in the same way that inflections do not generally develop idiosyncratic, unpredictable meanings when combined with specific lexical items):
“to hold close to”
kí:-ko-nam-ci kà:t la:kô:k
past3-3-hold-dat neck children
“(S/he) spoiled the children.”
“to sit on”
“to have a baby”
“(She’s) recently given birth.”
“to sleep with”
kó:-ay-ta kè:y pi:c-à:
past2-do-itv refl people-those
“Those people slept together.”
- The abstract shape of this suffix may be given as /-u/, but with infinitives there is only a tonal (and sometimes an [ATR]) contrast to be seen. With verbs which have /ke:-/ infinitival prefixes, the stem vowel tone is low (in contrast to high); with verbs which have a /ki:-/ infinitive, the stem vowel is high toned (in contrast to high-falling tone).
- This form is grammatical with the meaning ‘the dog is stuffed’ (i.e. has eaten so much that its stomach is severely bloated). Cf. /ke:-wi:rèn/ ‘to be bloated from overeating’.
- Cf. Jakobson 1990: 358: “...the D[ative] denotes the existence of its referent as independent of the action.”
- Cf. the Concise Oxford Dictionary’s example, “(of equipment, etc.) used with a computer but not an integral part of it”. In every instance, the added participant is involved in the action but not in a central way. Jakobson 1936 uses the term Randkasus (Rand = ‘edge, margin’) to characterize the Russian dative and instrumental cases as opposed to the nominative and accusative cases, for which he uses the term Vollkasus (Voll = ‘full’). In the English translation of his article: “A peripheral case indicates that its referent occupies a peripheral status in the overall semantic content of the utterance (1990: 352).”
Itive /-ta/ can, however, participate in diathetic alternations with underived forms:
“to throw at the dog”
“to throw the stone at”
“to throw the stone thither”
- Ashton, E. O. 1947 Swahili Grammar, 2nd ed., London, Longmans.
- Creider, J. T. and C. A. Creider 1989 A Grammar of Nandi, Hamburg, Buske.
- Creider, J. T. and C. A. Creider 2001 A Dictionary of the Nandi Language, Cologne, Rüdiger Köppe.
- Croft, W. 1991 Syntactic Categories and Grammatical Relations: the Cognitive Organisation of Information, University of Chicago Press.
- DeLancey, S. 2000 “The universal basis of case”, Logos and Language 1-2: 1-15.
- Fillmore, C. 1971 “The grammar of Hitting and Breaking”, in R. Jacobs and P. Rosenbaum, Readings in English Transformational Grammar, London, Ginn, p. 120-33.
- Fillmore, C. 1977 “The case for case reopened”, in P. Cole and J. M. Sadock, Syntax and Semantics 8: Grammatical Relations, New York, Academic Press, p. 59-81.
- Grimshaw, J. 1990 Argument Structure, Cambridge (Mass.), MIT Press.
- Jackendoff, R. 1972 Semantic Interpretation in Generative Grammar, Cambridge (Mass.), MIT Press.
- Jakobson, R. 1936 “Beitrag zur allgemeinen Kasuslehre”, Travaux du Cercle Linguistique de Prague 6: 240-88.
- Jakobson, R. 1990 “Contribution to the general theory of case”, in L. R. Waugh and M. Monville-Burston, On Language, Cambridge (Mass.), Harvard University Press.
- Larsen, I. 1979 Towards a system of semantics M.A. thesis, University of Nairobi.
- Levin, B. 1993 English Verb Classes and Alternations, University of Chicago Press.
- Lyons, J. 1977 Semantics, 2 vols., Cambridge University Press.
- Rosch, E. and C. B. Mervis 1996 “Family resemblances: studies in the internal structure of categories”, in H. Geirsson and M. Losonsky, Readings in Language and Mind, Cambridge (Mass.), Blackwell, p. 442-460.
- Rottland, F. 1982 Die Südnilotischen Sprachen: Beschreibung, Vergleichung und Rekonstruktion, Berlin, Dietrich Reimer.
- Smyth, H. W. 1956 Greek Grammar, 2nd ed. rev. by Gordon M. Messing, Cambridge (Mass.), Harvard University Press.
- Talmy, L. 1985 “Lexicalisation patterns: semantic structure in lexical forms”, in T. Shopen, Language Typology and Syntactic Description. Vol. 3. Grammatical Categories and the Lexicon, Cambridge University Press, p. 57-149.
- Tucker, A. N. 1964 “Kalenjin phonetics”, in D. Abercrombie, In Honour of Daniel Jones, London, Longmans, p. 445-470.
- Tucker, A. N. and M. A. Bryan 1966 Linguistic Analyses: The Non-Bantu Languages of North-Eastern Africa, London, Oxford University Press (for the International African Institute).