DKA

When does Kinyarwanda mean what it says? Towards a theory of sentence processing.

Kinyarwanda has different types of ambiguities that it is important to find out how different existing models of sentence processing can adequately handle the data presented by this language. In addition to ambiguities found in other languages namely lexical, structural, idiomatic and pragmatic, Kinyarwanda uses very productively hyperboles, bipolar expressions, euphemisms, taboo language, onomastic expressions, vulgar language and many de-semantacized expressions in the regular everyday speech, which contributes to this ambiguity. The lexical ambiguity is due to homonymy, polysemy, and the mixing of dialects and specialized languages vocabulary in everyday language. Besides lexical homonymy, other types of homonymy result also from juncture phenomena and from phonogical rules which create phonetic neutralizations, and from both morphological and syntactic rules which also create identical surface structures. The ambiguity is thus paradigmatic (lexical) and syntagmatic (structural).

The structural ambiguity is caused by decategorization or the hybrid nature of some words such as finite verbs which can function as either main verbs, auxiliairies or function words as complementizers, conjunctions or coordinators, and syntactic operations which create the same structural configurations with other unrelated syntactic structures such as relativization, object-subject reversal, multiple pronoun-incorporation.

Kinyarwanda speakers are aware of the multiple ambiguities that their mother tongue can create. Not only many anecdotes abound about people who did funny things because they only understood one meaning which was not intended by the speaker. The most common are jokes by and about Semabinga, a well known personality during the colonial period, who would anger or make authorities laugh by intentionally doing the opposite of what he was told to do or would respond in an unpredicted funny manner because of the utterance ambiguity.

There is also a word play, a linguistic duel, to be precise called gucyóocyoorana 'to tease each other' or amahuúngu 'boy talk' in which the interlocutor tries to outsmart the other person by responding to what he or she had just said in a such a way that the challenger feels hurt or humiliated. This duel stops when one of the interlocutors "loses" by failing to find an appropriate response.

Poetry is also there to test both real world and linguistic knowledge. The poet is not only interested in creating an object of verbal aesthetic beauty pleasing to the ears by its rhythms, rhymes, and a harmonious association between sound and sense but also in elevating the intellect of the listeners in helping them to understand the hidden meanings of language by using a medium that only the ones who have mastered the lexicon and the structure of language, the culture and the history of the country can understand. The poet is called umusízi 'the one who outruns others' and to compose poetry is called gusíga 'to distance/outrun others'. To understand the poem a process called gusíguura is performed. A proper translation of the latter term would be 'to decipher' or 'to decode'.

The Kinyarwanda data clearly show that to get a correct interpretation of natural language utterances and texts, a good theory of natural language processing is needed. Clearly both linguistic and non-linguistic contexts are in order. This research is of both theoretical and practical importance. It shows how the mind works and it can also contribute to the writing of software which have to do with machine translation. By using Peircian approach to the study of signs, an association, or similarity is seen among all these different types of ambiguities.

They all fall into three categories namely:

1. Paradigmatic and Syntagmatic Homonymies
2. Paradigmatic and Syntagmatic Polysemies
3. Deiconized or Grammaticized Paradigmatic and Syntagmatic Polysemies
   1. De-metaphorized ambiguities
   2. De-metonymized ambiguities

In this approach, a sign or structure can be considered as a paradigm at one level and a syntagm at another level just like in other systems. The semantic structure consists of several layers or tiers and it is not necessarily linear. In other words, semantic processing is represented the way phonological features and rules are represented in autosegmental and metrical phonology. It starts with the stem to the word, the phrase, the sentence, and the utterance and the whole text.

Kinyarwanda specialized languages

Kinyarwanda has one ethnic dialect only: Namely the one spoken by Batwa. This one is also disappearing because of the decrease of its speakers and the Tutsi genocide by a Hutu supremacist regime which took place in 1994 killing more than one million people within only three months space. The other ones are mostly regional: two main dialects the northern dialect igikiga with its subdialects (ikirera/ikigoyi) and the southern dialect (ikinyanduga) with its subdialects also namely the southwest dialect spoken in Kinyaga, the southeast dialect spoken in Gisaka and the south-central dialect spoken in Nduga. The differences among these dialects are mainly lexical and phonetic.

The specialized languages use the same type of vocabulary as the regular language but these vocabularies acquire different meanings. A careful study shows also an association or similarity between these acquired meanings with those of the regular language. Sometimes archaic or loan-words from different dialects or neighboring languages are also found. The exception is the language used in the initiation of Lyangombe (kubaandwa). It too borrows from the ordinary language but many times these words are phonetically deformed by either adding syllables, deleting them, inserting others, or by substituting sounds.

These specialized languages are: the language of poetry (ubusízi), the blacksmith language (urucúzi), the hunting language (amahiígi), the apiculture language (amahuúmvu), initiation language (imaándwa), coronation language (ubwíiru), divination language (kuragura), stone game language (igisoro), war dance language (guhamiriza), decoration language (gutaaka), cards game language (amakáratá) and hoop language (uruzíga). This type of specialized vocabulary is like professional jargon found in other languages.

There are three types of poetry in the language each with its own vocabulary namely the dynastic poetry (ubusízi) which praises kings, the panegyric poetry (ibyíivugo) which consists of praises for great warriors and national heroes and pastoral poetry (amazína y'íinká) which praises cows.

Homonymous and Polysemous ambiguity

The majority of ambiguities found in natural languages are due to either homonymy or polysemy. Homonyms are words which accidentally happen to have the same phonetic shape as illustrated by the examples in (1).

1

Polysemous expressions or tropes, on the other hand, are signs whose referents belong to the secondary plane of expression or words with "extended meanings" to use a more familiar terminology. Because of this very nature, tropes are ambiguous because it may not be clear as to whether they are indeed tropes or used in the primary plane of expression. There are two types of tropes: metaphors and metonymies. Tropes are metaphors if their referents have similarities either physical or functional with those of the primary plane (size, shape, color, function, ...).

2

They are metonymies if an association exists between the referent of the secondary plane and that of the primary plane such as cause and effect, content and container, possession and possessor, part and whole, work and artist, product and origin, etc.

3

Like English in which words such as wind [wind] and [wajnd], bow [baw] and [bow], tear [tijr] and [tejr], etc. are ambiguous because they are spelled the same way, Kinyarwanda also has homographic ambiguities due to the fact that the official orthography doesn't mark tones and vowel length as examples in (4) show:

4

In this paper this orthographic ambiguity will be ignored.

Polysemous or polyfunctional morphemes

Like regular words, smaller morphemes: function words, suffixes, and prefixes can also have multiple functions or meanings. The same noun class marker can represent nouns in different categories for instance and the same verbal suffix can have multiple functions as examples below show.

Noun class markers:

5

Below is a table with the noun class markers and their meanings split into two columns:

The associative morpheme -a

The associative morpheme has many functions: possession, qualification, identification, enumeration, accompaniment, origination, etc. For this reason it creates ambiguities.

6

Below is a table that separates each example into its original expression, two ambiguous interpretations, and the literal breakdown:

This is the same as the applicative suffix -ir- which is added to the verb stem to show different types of grammatical roles: benefactive, dative, possessor, goal, locative, etc.

7

The causative suffix -iish- can also stand for the cause and the instrument:

The clitics -hó, -mó, and -yó are found with many verbs to create idiomatic expressions or are used as pronouns referring to specific antecedents.

8

Many space and time concepts share the morphemes also resulting again in many ambiguities as we see below.

9

In the next sections, we will see that the ambiguities which are discussed also fall into these two categories: homophonous paradigmatic or syntagmatic ambiguities and polysemous paradigmatic or syntagmatic ambiguities because of the accidental similar phonetic shape or because of the tropical relationship with referents in the primary plane of expression.

Everyday language

In everyday language, different types of figure of speech can be used. These are idiomatic expressions, euphemisms, vulgar language, taboo language; hyperboles; names; empty words; auto-opposite words.

Idiomatic expressions.

As I have shown earlier, idioms are just like other tropes that is metaphors and metonymies or deiconzed tropes. The only difference between regular tropes and idiomatic expressions is that the former consist of single lexical items whereas the latter consist of larger structures or syntagms (Noun Phrases, Verb Phrases, Adjective Phrases, Adverbial Phrases, etc. ).

Euphemisms.

Euphemisms are nice or polite words which are used to refer to referents which are unpleasant to talk about such as death, bodily functions, etc. such as 'to pass away', 'sanitary napkins', 'restrooms', 'to make love', etc. in English. Here are some examples from Kinyarwanda.

10

Euphemistic terms are also used to refer to the king's activities and body parts and to cows as well. Illustrating examples are presented in 11 (a) and 11 (b), respectively.

11 (a) – King’s Euphemisms

11 (b) – Cow Euphemisms

The taboo language

The taboo language is used by women in what is known as gutsíinda . A woman can never pronounce the name of her parents-in-law. These are not restricted to the husband's father and mother but also to his uncles and their wives and to his aunts and their husbands. Words which sound like their names are also avoided and substitutes have to be found.

A full discussion of this linguistic phenomenon is discussed at great length in Kimenyi (1994).

12

Some specific taboo words also which take prefixes of all classes have been created to refer to any object (animate or inanimate): -seréza, -shemére, etc.

Many of the taboo language vocabulary are borrowed from the hunting language. The hunting language is used mostly to avoid bad luck such as being killed in the game or coming back from hunting empty-handed.

Although many taboo words don't have other referents in the primary plane, the majority are metonymies or metaphors also.

Hyperboles

Hyperboles are used to exaggerate. They are also either metaphors or metonymies: having similarity or the association between the referents of the primary plane and those of the secondary plane.

13

The last [4] expressions are used to complement somebody for what he or she has just done or the looks such as new hairstyle, new clothes, etc.

Thus to express a large quantity of anything the following expressions are used:

14

Example of usage:

• inzu yé ní u Rwanda/inzu yé ní amahaánga translates as "His/her house is very big"

• "Haaje abaantu ikivú/ubushyó/ishyano ryóose/urugaámba/isoko…" translates as "A lot of people came."

The last examples are metaphors, whereas the first ones are metonymies.

Vulgar language

The vulgar language is used only with close friends or with people that the speaker despises. It is never used with superiors or strangers. Below examples are provided:

15

The majority of these vulgar expressions refer to every day activities such as 'eat', 'drink', 'sit down', 'stand up', 'sleep', etc. and many synonyms for these referents are found as seen below.

16

The majority of the vulgar language vocabulary are metaphors. A physical or functional similarity is found between what is being referred to and the referent of the primary plane. Others don't have any other referent and belong to the primary plane only. Other factors, however, such as sound symbolism are utilized to convey the vulgar concept.

17

Bipolarity expressions

Kinyarwanda has many words and expressions whose meanings are found in the opposite poles, one being the opposite of each other such as to dust: 'to put dust'/ 'to remove dust'; to weed: 'to put weed'/ 'to remove weed'; sanction: 'to reward'/ 'to punish'; to rent: 'to be a landlord'/'to be a renter', etc. in English.

I refer to them as bipolarity expressions. They can be nouns, noun phrases, verb phrases, sentences, or simple morphemes. They are found mostly with derivative expressive morphemes of class 7, 8, 11, 12 and 14, the negative sentences, the hortative mood and ideophones.

Class markers -ki-, -bi-, -ru- are augmentative morphemes but are also both pejorative and ameliorative morphemes in both moral and physical sense.

18

The class marker -ka-, -tu- and -bu- are also diminutive morphemes but they too have also an ameliotive or pejorative meaning.

19

20

21

This scope ambiguity is possible however only if the quantifier is in the head noun domain:

22

The hortative mood with intonation or repetition, produces the same effect as the negative discussed above:

23

The morpheme -ye is usually referred to as the perfective aspect marker because it shows a completed action or a state. In some verbs, however, it stands for both perfective and imperfective:

24

Praise expressions have both opposite meanings also:

25

Below is a table summarizing the praise expressions and their opposite meanings:

Many ideophones and interjections have bipolar meanings as well. The most common meanings are pain and pleasure, excitement and disappointment, approval and disapproval.

26

Some words have the opposite of what they mean depending on the subject or the object. The word gupfá 'to die' and kwíica 'to kill' are some of those. With the word ibyoobo/ibihunnyo 'gupfá' means to 'grow' and kwíica 'pick':

27

Irony which is found in many languages is a figure of speech also in which the form is intented to express the opposite. Here are again some examples from Kinyarwanda:

28

Some verbs with the causative suffix -iish- or -y- can also be ambiguous as far as bipolaority is concerned:

29 (a)

29 (b)

29 (c)

Clearly to get the right interpretation of bipolar expressions, extralinguistic (pragmatic) information is needed.

Onomastics

Kinyarwanda names appear in all grammatical structures as verbs, nouns, noun phrases, verb phrases, sentences, etc.and are semantically transparent (Kimenyi, 1979). There are times therefore when it is not possible to tell if they are regular utterance constituents or names.

30

Secondly, each animate object (humans and animals) and inanimate can have surnames which look exactly like regular personal names. These surnames are iconic since they describe the object's attributes such as physical shape, moral character or function. Simple names are obtained by deleting from any word the preprefix or by adding to it onomastic prefixes such as nyir-, nyirá-, mukáa-, sée-, rwáa-, káa-, etc.

31

There are regular words which also look like personal names. These refer mostly to types of crops, plants, insects, birds and animals, events:

32

The abundance of surnames in Kinyarwanda originates probably from praise-poetry. Humoristic folk praise-poetry has indeed developed from it where funny objects are given praise-names even praise-poems just for entertainment purposes.

Ambiguities created by phonetic neutralization

Phonetically motivated ambiguities are due to phonological rules which neutralize individual phonemes into the same phonetic realization in certain predictable environments such as the neutralization of t and d in English into a flapped voiced alveolar [rajDer] for rider or writer.

The following Kinyarwanda examples show this type of ambiguity

33

Tone neutralization caused by the tense-aspect-modality especially in subordinate and negative clauses is also responsible of this type of ambiguity.

34

Many verbs with contrasting lexical tones are also neutralized in these tenses.

Thus, verbs such as kuvuna 'to come to the rescue' and kuvúna 'to break'; gutoonda 'to lose recollection' and gutóonda 'to climb'; kuyoomba 'to finish very slowly/to disappear very slowly and kuyóomba 'to make a slow rhythmic sound'; etc. are neutralized in many negative and subordinate tense-aspect-modality:

35

Obviously, phonetically motivated ambiguities are paradigmatic homonyms.

Juncture phenomena, reanalysis, or folk etymology

This type of ambiguity is due to morphemic segmentation or the assignment of boundaries.
Sometimes the word is wrongly segmented, giving another meaning than the intended one.
This occurs in both the prefix and suffix position in which the stem is shortened by detaching the segments on the right or the segments on left thought of as being prefixes or suffixes. This occurs mostly in verb forms.

36

This type of ambiguity is similar to the English juncture phenomena: a notion/an ocean; a night rate/a nitrate; I scream/ice cream; an aim/a name; saturday/sadder day; an ice man/a nice man; get aboard/ get a board; great ape/grey tape; peace talks/pea stals; keep sticking/keeps ticking...

Structural or Syntagmatic Ambiguity

This section discusses ambiguities found in larger grammatical constituencies. The ones to be discusses are serial verb construction, relative clauses, object-subject reversal and multiple pronoun-incorporation.

Ambiguity due to verb serialization

Function words

Function words in Kinyarwanda namely conjunctions, subordinators, etc. derive from regular verbs or look exactly like regular verbs. This phenomenon therefore causes ambiguities at the clause or sentence level.

38

Auxiliaries

Auxiliaries are also regular verbs, what makes them auxiliary is their position in the sentence structure. I have refered to this as syntagmatic or structural derivation.

39

Relative clauses

There is no segmental marker to mark the relative pronoun. The relative clause verb is marked by a high tone on the nucleus of the second syllable of the verb stem if the antecedent is an object but on the nucleus of the first syllable if the antecedent is subject. The lack of a relative pronoun makes relative clauses ambiguous.

40

Object-Subject Reversal

Object-subject reversal consists of interchanging the subject and the object. The former object acquires all the subject properties namely verb agreement. The new sentence has a passive meaning, however. This is discussed in great detail in Kimenyi (1980).

41

When both subjects and objects are animate, however, the ambiguity is created.

42

Multiple pronoun incorporation

Like Romance languages, Kinyarwanda allows multiple object clitics into the verb. They all precede the verb stem. Like in Romance languages also, there is a strict word order. This strict order depends either on the case role hierarchy or on the person hierarchy. Thus when it is the case hierarchy, the benefactive comes close to the verb stem, followed by the dative and farther away the accusative. The only case which is not bound by this constraint is the locative which can "float" ( for more details see Kimenyi, 1980). This is illustrated in (1)

43

In the person hierarchy, it is the reflexive morpheme (-ii-) which comes next to verb stem, followed by the first person singular or plural , followed by the second person (singular or plural) and then the third person:

44

Incorported object pronouns of causative constructions cause ambiguities as well.
When both the causee and the causer clitics appear in the same verb, there is a fixed word order also. The object of the extension always comes next to the verb stem.

45

There is no way the ambiguity can be avoided especially when both object clitics belong to the same category (human/animate/inanimate).

Desemanticized expressions

Many words and expressions without any semantic content are also used in everyday language. These happen to be words or expressions which refer to greetings, exclamations, interjections, swears, dicourse fillers, surprise. These ofcourse have meanings in other linguistic contexts.

Expressions showing surprise

Surprise or exclamations are used when one sees something exciting or something that one were not expected to see.

46

These would correspond to English expressions such as Oh! my God! Jesus Christ!

Addressee expressions

The following expressions are used mostly when the speaker is asking him/herself questions outloud as if s/he was expecting an answer or response to the real or imagined listener.

47

Again these expressions can be translated in English by some of the following: brother! man! boy! my God.

Expressions referring to greetings

Many of the greetings are about good wishes for wealth, health, and longetivity.
Others, however, out of an appropriate context, can be ambiguous:
For instance, Nimuduhé! 'Give us' is used when entering somebody's house-compound to announce oneself. Ngaahó turakóra. 'You see us working!' is the response.
Muraaré aharyáana 'spend the night where (the body) itches' is equivalent to Goodnight! in English. In other words, it means exactly the same as the other expression which is used namely uraramuke(hó) 'survive the night!'. The body which doesn't itch is a dead body.

Polite expressions used to leave the interlocutor

These desemanticized or conventionalized expressions are illustrated by the following examples:

48

Rwandan husbands don't beat their wives and the wives cannot kick their husbands out since the husband is the head of the household and is the one who can repudiate his wife.

All these words and expressions, even though they are now devoid of semantic content, were at once tropes. But because of cultural change and linguistic dynamism, they have reached the level of the symbolic sign: the linguistic opacity and lack of relation between the linguistic sign and what it stands for. For instance, many expressions referring to the king especially in greetings, exclamations and swearing are still used in the language eventhough the country has been a republic for the last four decades.

Conclusion

A complete adequate grammar of natural languages has to generate and process, at the same time, acceptable utterances. It has to provide both the code and the decoding simultaneously. Unfortunately, for the last forty years, mainstream linguistics (generative grammar) has mostly focused on the former (sentence generation) sacrificing the latter (sentence processing). The latter seems to be more challenging, however, and more interesting because it is the one which closely deals with conceptual structures. In other words, how does the mind recognize the intended meanings when it is confronted with many competing semantic interpretations? The grammar of utterance production is indeed less complex when compared to the decoding aspect. There is indeed a small number of grammatical structures available and the rules which create them consist of a finite set of numbers as well. There is no limit, on the other hand, on what the language can express.
The human mind can be indeed understood metaphorically as a super computer where the brain is the hardware and the mind a set of of various softwares in which language is generated and precessed by its own software. These cognitive softwares (language, memory, vision, etc.) have to work simultaneously. Obviously, there must exist a central software probably the mind which does the coordination. In this approach any single linguistic item is considered as being a paradigm and a syntagm at the same time. A large structure is also viewed as being a paradigm as well. The way for instance, a linguistics department might be a paradigm when it is a university unit but a syntagm when it is a whole with its faculty and program. Unlike traditional generative grammar, parsing doesn't come after the utterance has been completed: that is from the top to the botton: sentence boundary, major constituents boundaries (NPs, VPs, etc) to small constituents. In this approach, processing starts as soon as speech starts and ends immediately when it stops. Each paradim and syntagm is also assigned all possible meanings and functions and identifies them as either homonymous or polysemous. If they are polysemous then, the parser has to tell whether they are metaphors or metonymies. This investigation into language processing has also shown that semantics cannot divorce itself from sociolinguistics. The language software or the dictionary has to recognize all the different dialects, specialized languages or jargons and the different registers for proper semantic intepretation. Communication is easier in societies with centralized governments because there exist language policies and the imposition of prescriptive grammars. A lot of cliches are thus used as opposed to societies which are not nations yet.

References

• Kimenyi, Alexandre 1989. Kinyarwanda and Kirundi names: a semiolinguistic analysis of Bantu onomastics.
• New York: The Edwin Mellen Press. 1980. A relational grammar of Kinyarwanda.
• Berkeley and Los Angeles: University of California Press. 1993. "Why is it that women in Rwanda cannot "marry"? in Locating power Berkeley Women Linguistic Society.