Mova0.6 pre-release

the flag of India 17. Derivation

Derivation forms a lexeme from an existing base. A derived word can name a participant, an activity, a quality or a new kind of event. It can also change word class, as when a verb yields an activity noun, or change the participants that a verb can take, as in a causative. Kannada uses suffixes extensively for these purposes. It also has prefixes in learned vocabulary and prefix-like first elements in some word families.

Roots, bases and productive affixes

A root is the smallest lexical element identifiable in a word family. A base is whatever a particular operation applies to. It can be a root, a stem with a characteristic final vowel, an already derived stem or a compound. A stem supplies the form to which inflection attaches. Thus in ಕೆಲಸಗಾರರು kelasagāraru ‘workers’, the noun ಕೆಲಸ kelasa ‘work’ is the base for the formation of ಕೆಲಸಗಾರ kelasagāra ‘worker’, and that derived noun supplies the stem for the plural. The Kannada term samāsa denotes a compound, which contains more than one lexical member and can itself serve as a base (Wikibooks contributors 2026, “Compound bases”).

Form and romanizationGlossTranslation
ಕೆಲಸ kelasawork‘work’ (constructed)
ಕೆಲಸಗಾರ kelasa-gārawork-AGENT‘worker’ (constructed)
ಕೆಲಸಗಾರರು kelasa-gāra-ruwork-AGENT-PL.HUM‘workers’ (constructed)
ಕೆಲಸಗಾರರಿಗೆ kelasa-gāra-r-igework-AGENT-PL.HUM-DAT‘to the workers’ (constructed)

The suffix -gāra gives the word a human referent and creates a noun that can be used independently. Plural and dative marking locate that referent in a sentence. They apply to the derived noun, and the dative follows the plural when both are present. The segmentation -r-ige shows the plural stem before the dative in this example. In the glosses, AGENT means an agentive suffix, PL.HUM means human plural and DAT means dative.

A productive pattern is available for forming new words on eligible bases. Productivity depends on the pattern and the sense the speaker intends. A speaker can interpret an agent noun in -gāra when its base names a recognizable occupation or pursuit. An existing word can supply a more usual name for the same person, and an established derivative can acquire a narrower sense than its parts predict. A recurrent ending in a set of established words establishes a pattern, while the ease of extending it to a new base is a separate property.

An affix can be followed by other material without being an infix. In a form built from a verbal base, a causative suffix and a tense ending, the causative still follows the base. An infix would interrupt the base itself. A root-and-pattern system would form a family by fitting root material into a recurrent internal template. The Kannada formations illustrated here use concatenation and stem alternation. Internal differences within a word family, including the vowel differences in some learned forms, are treated with the particular formation that licenses them.

Deriving agent, abstract and other nouns

Agents and people associated with a base

Human nouns can identify a person by an activity, occupation, skill or affiliation. The base and suffix together constrain the interpretation. In occupational nouns, -gāra names someone who performs the activity or is associated with it as work. The paired form -gārti can specify a woman in families that have this pair. In affiliation nouns, -iga can identify a person by a language or community. Some action nouns for people have -ga after a vowel-final verbal base. These are separate patterns, and the choice of ending for an established name belongs to the word family. Kittel indexes iga among the taddhita formations (Kittel 1903, p. 462).

Base and glossDerived form and romanizationGlossTranslation
ಕೆಲಸ kelasa ‘work’ಕೆಲಸಗಾರ kelasa-gārawork-AGENT‘worker’ (constructed)
ಕೆಲಸ kelasa ‘work’ಕೆಲಸಗಾರ್ತಿ kelasa-gārtiwork-AGENT.F‘woman worker’ (constructed)
ಆಟ āṭa ‘play, game’ಆಟಗಾರ āṭa-gāragame-AGENT‘player’ (constructed)
ಆಟ āṭa ‘play, game’ಆಟಗಾರ್ತಿ āṭa-gārtigame-AGENT.F‘woman player’ (constructed)
ಕನ್ನಡ kannaḍa ‘Kannada’ಕನ್ನಡಿಗ kannaḍ-igaKannada-AFFILIATE‘Kannada speaker’ (constructed)
ಓದು ōdu ‘read’ಓದುಗ ōdu-garead-AGENT‘reader’ (constructed)
ಕೇಳು kēḷu ‘hear, listen’ಕೇಳುಗ kēḷu-galisten-AGENT‘listener’ (constructed)

The verb ಓದು ōdu also occurs as a noun meaning ‘reading’ or ‘study’. The agent noun ಓದುಗ has its own lexical use. Regular nominal endings apply after agent formation, as in the number and case forms of ಕೆಲಸಗಾರ above. Human plural marking can apply to nouns with an overt feminine derivative as well, as in ಆಟಗಾರ್ತಿಯರು āṭagārti-y-aru ‘women players’, glossed game-AGENT.F-LNK-PL.HUM (constructed). LNK marks a linking sound at a boundary. Learned agent names can have their own gender pairs. ಲೇಖಕ lēkhaka ‘author’ and ಲೇಖಕಿ lēkhaki ‘woman author’ illustrate such a pair (constructed). The two words share a learned stem associated with writing, and the pair provides a model within that lexical stratum.

Qualities, states and activities

-tana forms nouns for a quality, condition or characteristic behavior. A property base gives a quality reading. A base naming a kind of person can give the condition or conduct associated with that person. -te and -tva are frequent in learned quality nouns. Their distribution is lexical in established vocabulary, so consult the individual family for both the ending and the sense. An abstract noun can acquire a more specific conventional meaning, such as a term for an institutional kind of equality.

Base and glossDerived form and romanizationGlossTranslation
ಒಳ್ಳೆಯ oḷḷeya ‘good’ಒಳ್ಳೆಯತನ oḷḷeya-tanagood-QUALITY‘goodness’ (constructed)
ಹುಡುಗ huḍuga ‘boy’ಹುಡುಗತನ huḍuga-tanaboy-QUALITY‘boyhood, boyishness’ (constructed)
ಮೂರ್ಖ mūrkha ‘foolish, fool’ಮೂರ್ಖತನ mūrkha-tanafoolish-QUALITY‘foolishness’ (constructed)
ಸಮಾನ samāna ‘equal’ಸಮಾನತೆ samāna-teequal-QUALITY‘equality’ (constructed)
ಸರಳ saraḷa ‘simple’ಸರಳತೆ saraḷa-tesimple-QUALITY‘simplicity’ (constructed)
ಮಾನವ mānava ‘human’ಮಾನವತ್ವ mānava-tvahuman-QUALITY‘humanity, humaneness’ (constructed)

A verbal base can yield a noun for an act, result or state. Forms in -ike illustrate the range. The form and established sense both matter. A noun for praise can name the act or its content, and a noun for belief can denote a relatively lasting state. Kittel's index groups ike with ikke and lists ige in the same derivational use. It separately records similar forms in the case system. The ending alone therefore leaves the analysis open until the base and the use of the whole word have been identified (Kittel 1903, p. 462).

Verbal base and glossDerived form and romanizationGlossTranslation
ಹೊಗಳು hogaḷu ‘praise’ಹೊಗಳಿಕೆ hogaḷ-ikepraise-NMLZ‘praise, commendation’ (constructed)
ನಂಬು nambu ‘believe, trust’ನಂಬಿಕೆ namb-iketrust-NMLZ‘belief, trust’ (constructed)
ಓಡು ōḍu ‘run’ಓಟ ōṭarun.N‘a run, running’ (constructed)
ನೋಡು nōḍu ‘see, look’ನೋಟ nōṭalook.N‘a look, a view’ (constructed)

NMLZ marks nominalization. In the first two pairs the final u of the cited verb gives way to the vowel-initial suffix. The last two pairs have a conventional noun stem with ṭ where the verb has ḍ. Their meanings and stem shapes belong to their respective families.

A clause headed by a non-finite verb can also refer to an activity. In ಪುಸ್ತಕ ಓದುವುದು pustaka ōd-uv-udu ‘reading a book’, glossed book read-NPST-NMLZ (constructed), ಪುಸ್ತಕ pustaka ‘book’ is still the object of ‘read’. That construction is available when the speaker wants to refer to an event together with its participants. A lexical derivative such as ಓದುಗ ōduga ‘reader’ is a noun referring to a person and takes the ordinary inflection of that noun class.

Deriving adjectives and adverbs

Property words and attributive forms

A suffix can describe something by the quality, substance or domain named by the base. -vaṃta characterizes someone as possessing a quality or resource and is common in descriptions of people. -maya characterizes something as filled with, pervaded by or made up of what the base names. Learned formations in -ika often express a relation to a domain. The surface relation to the corresponding noun can involve changes in the base as well as a suffix.

Base and glossDerived form and romanizationGlossTranslation
ಬುದ್ಧಿ buddhi ‘intelligence’ಬುದ್ಧಿವಂತ buddhi-vaṃtaintelligence-POSSESSING‘intelligent’ (constructed)
ಧನ dhana ‘wealth’ಧನವಂತ dhana-vaṃtawealth-POSSESSING‘wealthy’ (constructed)
ಆನಂದ ānanda ‘joy’ಆನಂದಮಯ ānanda-mayajoy-FILLED‘full of joy’ (constructed)
ಸಮಾಜ samāja ‘society’ಸಾಮಾಜಿಕ sāmāj-ikasociety-REL‘social’ (constructed)
ದೇಹ dēha ‘body’ದೈಹಿಕ daih-ikabody-REL‘bodily, physical’ (constructed)

In attributive use a property word precedes its head noun. Some learned descriptions of people also have substantival uses. For example, ಬುದ್ಧಿವಂತ can serve as an attribute or, in a context with a human referent, as a masculine noun meaning ‘an intelligent man’. Kannada also builds attributes from inflected nouns and from relative participles. Kittel's index identifies the genitive use of a separately from its derivational use and lists āda as a past relative participle of āgu ‘become’ (Kittel 1903, pp. 458, 461). In a modern phrase in -vāda, the construction with āda can express a property of the head noun without an English past-tense translation.

Form and romanizationGlossTranslation
ಬುದ್ಧಿವಂತ ಹುಡುಗ buddhi-vaṃta huḍugaintelligence-POSSESSING boy‘an intelligent boy’ (constructed)
ಸಾಮಾಜಿಕ ಜೀವನ sāmāj-ika jīvanasociety-REL life‘social life’ (constructed)
ಮರದ ಮೇಜು mara-d-a mējuwood-LNK-GEN table‘a wooden table’ (constructed)
ಶಾಂತವಾದ ಸ್ಥಳ śānta-v-āda sthaḷacalm-LNK-become.PST.REL place‘a quiet place’ (constructed)

The genitive in ಮರದ ಮೇಜು expresses material in this phrase. Its noun remains a genitive form, even though English uses the adjective ‘wooden’. In ಶಾಂತವಾದ ಸ್ಥಳ the property expression contains a relative form. These two ways of supplying an attribute coexist with lexical adjective formation.

Manner and capacity

The formation in -āgi is related to the non-finite form of ಆಗು āgu. With a property base it can give a manner expression. With a noun denoting a role, status or capacity, it can mean ‘as’ or ‘in the capacity of’. The predicate and the meaning of the base determine the reading. A linking consonant can occur between the base and āgi. The forms below illustrate v after an a-final base and y after an i-final base. Kittel's index also points to treatment of āgi and records a separate adverb-forming use of an ending indexed as age (Kittel 1903, pp. 458, 461).

Form and romanizationGlossTranslation
ಸ್ಪಷ್ಟವಾಗಿ ಹೇಳು spaṣṭa-v-āgi hēḷuclear-LNK-ADV say.IMP‘Say it clearly.’ (constructed)
ನಿಧಾನವಾಗಿ ನಡೆ nidhāna-v-āgi naḍeslowness-LNK-ADV walk.IMP‘Walk slowly.’ (constructed)
ಅತಿಥಿಯಾಗಿ ಬಾ atithi-y-āgi bāguest-LNK-as come.IMP‘Come as a guest.’ (constructed)
ಬೇಗ ಬಾ bēga bāsoon come.IMP‘Come soon.’ (constructed)

Words such as ಬೇಗ bēga can be used adverbially in their lexical form. The -āgi construction is available when a property or capacity supplies the relevant meaning. A form ending in āgi can also retain the event sense ‘having become’ where the surrounding clause describes a change of state.

Verb formation and causative suffixes

Verbs from nominal bases

The verbal ending conventionally cited as -isu forms verbs on some noun and property bases, including many learned loan stems. The connection to the base can be ‘engage in the activity’, ‘direct the attitude or action toward someone’ or another established relation. The derived verb determines its own argument structure. A travel verb can have a traveler as its sole core participant, while a verb of respect can take a person as object.

Base and glossDerived form and romanizationGlossTranslation
ಪ್ರಯಾಣ prayāṇa ‘journey’ಪ್ರಯಾಣಿಸು prayāṇ-isujourney-VBLZ‘travel’ (constructed)
ಗೌರವ gaurava ‘respect’ಗೌರವಿಸು gaurav-isurespect-VBLZ‘show respect to, honor’ (constructed)
ಸಂತೋಷ saṃtōṣa ‘joy’ಸಂತೋಷಿಸು saṃtōṣ-isujoy-VBLZ‘rejoice, be glad’ (constructed)

The displayed bases drop final a before the vowel-initial ending. VBLZ marks the formation of a verb. A separate way to form a verbal predicate uses a lexical noun or property expression with a verb such as ಮಾಡು māḍu ‘do, make’ or ಆಗು āgu ‘become’. ಪ್ರಯಾಣ ಮಾಡು prayāṇa māḍu, glossed journey do, means ‘travel’ (constructed). Its lexical material is distributed over two words, and the verb carries the inflection. The single-word ಪ್ರಯಾಣಿಸು supplies a derived stem that takes verbal inflection directly.

Causatives

A causative introduces someone or something that brings about the event described by the base verb. A causative of an intransitive verb has a causer and a participant who undergoes the change or performs the action. A causative of a transitive verb can involve the causer, the person who performs the induced action and the original object. The verb and the intended relation between the participants affect their case marking. A context can leave an understood participant unexpressed.

The ending -isu is a common causative formation. The paired forms show how each base appears before it. Kittel's index groups historical forms with isu and directs the reader to the same pages for those variants (Kittel 1903, pp. 460, 462). The modern lexical inventory includes distinct words for drinking, sleeping, reading and other activities that provide bases for causatives (Wikibooks contributors 2026, “Verb”).

Base and glossCausative and romanizationGlossTranslation
ಮಲಗು malagu ‘lie down, sleep’ಮಲಗಿಸು malag-isusleep-CAUS‘put to bed, put to sleep’ (constructed)
ಓದು ōdu ‘read, study’ಓದಿಸು ōd-isuread-CAUS‘have someone read, have someone study’ (constructed)
ಕುಡಿ kuḍi ‘drink’ಕುಡಿಸು kuḍ-isudrink-CAUS‘give someone something to drink’ (constructed)
ಕಲಿ kali ‘learn’ಕಲಿಸು kal-isulearn-CAUS‘teach’ (constructed)
ನಡೆ naḍe ‘walk, proceed’ನಡೆಸು naḍe-suproceed-CAUS‘conduct, run, cause to proceed’ (constructed)

In the displayed u-final pairs, u is absent before -isu. The i-final pairs have a single i in the derived form. ನಡೆಸು has e before -su. These alignments describe the given stems. The meaning can also become conventional within a family. For instance, ಕಲಿಸು ‘teach’ can take a subject matter, and ನಡೆಸು ‘conduct’ can describe running a meeting or an institution.

Form and romanizationGlossTranslation
ಮಗು ಮಲಗಿತು. magu malag-itu.child sleep-PST.3SG.N‘The child fell asleep.’ (constructed)
ಮೀನಾ ಮಗುವನ್ನು ಮಲಗಿಸಿದಳು. mīnā magu-v-annu malag-is-id-aḷu.Meena child-LNK-ACC sleep-CAUS-PST-3SG.F‘Meena put the child to bed.’ (constructed)
ಮಗು ಹಾಲು ಕುಡಿಯಿತು. magu hālu kuḍi-y-itu.child milk drink-LNK-PST.3SG.N‘The child drank milk.’ (constructed)
ಮೀನಾ ಮಗುವಿಗೆ ಹಾಲು ಕುಡಿಸಿದಳು. mīnā magu-v-ige hālu kuḍ-is-id-aḷu.Meena child-LNK-DAT milk drink-CAUS-PST-3SG.F‘Meena gave the child milk to drink.’ (constructed)

The child is the subject in the simplex clauses. It is accusative in the clause about putting the child to bed and dative in the clause about giving the child milk to drink. The causer controls finite agreement in both causatives. The segmentation malag-is-id-aḷu places causative formation inside the tense and agreement morphology. A verb such as ಪ್ರಯಾಣಿಸು ‘travel’ has a similar outer shape, but its noun base and its single traveler argument support a verbalizing analysis.

Learned prefixes and inherited suffixes

Learned prefixes and loans

Sanskrit and Prakrit have contributed to Kannada vocabulary in different ways. A tatsama is a learned borrowing that preserves a form close to the source. A tadbhava is a historically adapted form whose sound shape has developed in transmission into Kannada. The loan history of a word and its present-day structure are separate properties. Kannada can inflect and derive words on borrowed bases, and some learned families contain recognizable prefixes or suffixes (Wikibooks contributors 2026, “Influence of Sanskrit and Prakrit”).

Kannada form and romanizationGlossTranslation and relation
ಬಣ್ಣ baṇṇacolor‘color’, an adapted form compared with Prakrit vaṇṇa and Sanskrit varṇa (Wikibooks contributors 2026, “Influence of Sanskrit and Prakrit”)
ರಾಯ rāyaking‘king’, an adapted form compared with rāja (Wikibooks contributors 2026, “Influence of Sanskrit and Prakrit”)
ದಿನ dinaday‘day’, a learned borrowing (Wikibooks contributors 2026, “Influence of Sanskrit and Prakrit”)
ಮುಖ mukhaface‘face’, a learned borrowing (Wikibooks contributors 2026, “Influence of Sanskrit and Prakrit”)

Learned prefixes occur chiefly in vocabulary built on learned bases. The privative pair a- and an- shows a consonant versus vowel environment in transparent pairs of this stratum. The other prefixes illustrated below contribute favorable evaluation, unfavorable evaluation or secondary status. Their meanings can be specialized in the resulting lexeme. A prefix-like element can also have a wider range of Kannada formations, as maru- does in words for repetition or renewal.

Element and useForm and romanizationGlossTranslation
ಅ- a-, privative before a consonant in a learned pairಅಶುದ್ಧ a-śuddhaPRIV-pure‘impure’ (constructed)
ಅನ- an-, privative before a vowel in a learned pairಅನರ್ಹ an-arhaPRIV-qualified‘unqualified, ineligible’ (constructed)
ಸು- su-, favorable evaluationಸುಗುಣ su-guṇagood-quality‘virtue, good quality’ (constructed)
ದುರ್- dur-, unfavorable evaluation in this familyದುರ್ಗುಣ dur-guṇabad-quality‘vice, bad quality’ (constructed)
ಉಪ- upa-, subordinate or secondary statusಉಪನಗರ upa-nagarasecondary-city‘suburb’ (constructed)
ಮರು- maru-, repetitionಮರುಪರಿಶೀಲನೆ maru-pariśīlaneagain-review‘re-examination’ (constructed)

Privative formations such as ಅಶುದ್ಧ are lexical property words. The prefix has scope over its base. Clause negation is a separate construction with scope over a predicate or proposition. The examples with su- and dur- illustrate a learned lexical pair on the same base. Familiarity with such a family helps a reader identify the element in other words, while the sense of each established formation can depend on its own use.

Inherited suffixes and traditional classification

The inherited suffixal stock is visible alongside learned morphology. Traditional descriptions use the label taddhita for a class of secondary formations. Kittel's index includes ame, avu, ike, iga and ika under this heading or through cross-references. It also lists a derivational a alongside several independently identified grammatical uses of the same vowel (Kittel 1903, pp. 457–462). The traditional classification covers historical material across Kannada varieties. Modern suffix choice, register and productivity have to be established for each family.

The difference between a borrowed suffix, a whole borrowed word and an inherited suffix on a borrowed base matters for comparison. For example, the Kannada family in -tana includes both inherited and learned vocabulary among its bases. ಮೂರ್ಖತನ mūrkhatana has the same transparent quality formation as ಒಳ್ಳೆಯತನ oḷḷeyatana. In ಸಾಮಾಜಿಕ sāmājika, the learned relation includes a change in the base vowel. The derivational category can be clear even when the history of an individual form includes both borrowing and subsequent extension within Kannada.

Productivity and stem alternations

A useful analysis begins with a relation in both meaning and form. The derived lexeme supplies evidence from its word class and inflection, while the base supplies evidence from independently occurring words or a recognizable family. ಕೆಲಸಗಾರ and ಕೆಲಸಗಾರರು show the agent formation followed by number marking. ಗೌರವಿಸು and its finite forms show a verb derived from a noun. In both families the lexical relation survives when the derivative is inflected.

Stem shape depends on the derivational pattern. Several recurrent relations are illustrated below. The segmented forms show the material at the derivational boundary. The last two rows show families whose stem changes are part of the lexical correspondence.

Stem relationForms and romanizationGloss and translation
The base vowel stays before a consonant-initial suffix in these formations.ಕೆಲಸ kelasa → ಕೆಲಸಗಾರ kelasa-gārawork → work-AGENT, ‘work’ → ‘worker’ (constructed)
Final a is absent before -iga in this affiliation noun.ಕನ್ನಡ kannaḍa → ಕನ್ನಡಿಗ kannaḍ-igaKannada → Kannada-AFFILIATE, ‘Kannada’ → ‘Kannada speaker’ (constructed)
Final u is absent before -ike in this noun.ನಂಬು nambu → ನಂಬಿಕೆ namb-iketrust → trust-NMLZ, ‘trust’ (verb) → ‘trust’ (noun) (constructed)
Final u is absent before -isu in this causative.ಮಲಗು malagu → ಮಲಗಿಸು malag-isusleep → sleep-CAUS, ‘sleep’ → ‘put to sleep’ (constructed)
A linking consonant occurs before -āgi in this manner expression.ಸ್ಪಷ್ಟ spaṣṭa → ಸ್ಪಷ್ಟವಾಗಿ spaṣṭa-v-āgiclear → clear-LNK-ADV, ‘clear’ → ‘clearly’ (constructed)
The lexical noun has ṭ corresponding to ḍ in the verb.ಓಡು ōḍu → ಓಟ ōṭarun → run.N, ‘run’ (verb) → ‘a run’ (constructed)
The learned adjective selects a stem with ai.ದೇಹ dēha → ದೈಹಿಕ daih-ikabody → body-REL, ‘body’ → ‘bodily’ (constructed)

Productive suffixation can coexist with a stem alternation. Once a speaker chooses a pattern, its base shape can be predictable for a group of words. The choice of that pattern can still be lexical. For quality nouns, the transparent behavior of -tana allows extension to new bases with an appropriate interpretation. Learned nouns in -te or -tva can be conventional for particular properties, especially in formal vocabulary. In verbal families, -isu is available for causative formations, while particular causatives can select a stem shape such as naḍe- or develop a specialized meaning such as ‘conduct’.

Identical letters at the end of a word can represent different functions. In a noun such as ನಂಬಿಕೆ nambike, the derived meaning and noun inflection identify a lexical noun. In a phrase such as ಕೆಲಸಕ್ಕೆ kelasa-kke ‘for work’, glossed work-DAT (constructed), the case ending sets the relation of an existing noun to the rest of the sentence. Kittel's separate derivational and case entries for ike, ikke and ige document the importance of this distinction in historical forms as well (Kittel 1903, p. 462). Likewise, a borrowed vowel alternation such as dēh- beside daih- belongs to its learned family. It supplies no general Kannada rule for replacing vowels inside new roots.

How Mova handles Kannada derivation

At level 1, tools/langseg labels sentences and spans. It assigns Kannada script to other. At level 2, tools/mova-parse uses a model for the language of each span. Kannada (kn) has no level-2 parser model. The parser leaves Kannada tokens with _ in LEMMA, UPOS, XPOS, FEATS, HEAD and DEPREL, and with Parse=None in MISC. Where a language model is missing, the parser reports that in the # parse comment. Mova has no annotated Kannada corpus and has measured no Kannada treebank baseline.

Squid reads the grammatical points in this chapter as claims. Formalizable claims, such as suffix rules and conditions on stem alternations, enter its queue. None of the Kannada derivation rules in this chapter is implemented. Mova currently supplies no Kannada tokenization, tagging, lemmatization or dependency parsing for these forms. Its frames and pragmatics read English trees and have no Kannada trees to read. The first derivational claims for implementation would be those needed by a tokenizer, a tagger and a Universal Dependencies treebank baseline.

Sources for this chapter

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/languages/kn/17-derivation/