Suppletion and defective paradigms in translation are two different ways in which real morphological systems depart from the tidy expectation that every lexeme has one stable stem and a complete set of predictable forms. Suppletion uses radically different forms inside one paradigm, as in go/went or good/better. Defectiveness leaves expected paradigm cells missing or restricted, forcing speakers to avoid them or use periphrastic alternatives.
People searching for suppletion, suppletive forms, defective paradigms, defective verbs, missing inflectional forms, irregular paradigms, go went morphology are solving a paradigm-identity problem. Current 2026 morphology research treats suppletion, syncretism, deponency and defectiveness as distinct departures from a canonical lexeme. Translation precision therefore requires identifying the lexeme and grammatical cell first, rather than inventing a regular form or treating an irregular stem as a new word.
This article owns the suppletion and defectiveness precision layer for eduKateSG. It does not replace Case Syncretism or general Word Formation. Its narrow task is to preserve one-lexeme identity across radically different stems, recognise genuine missing or restricted forms, use periphrasis where the target grammar requires it, and stop AI systems from regularising paradigms into forms that speakers do not actually use.
Quick Read
For every unusual form write LEXEME, PARADIGM CELL, SURFACE FORM and STATUS=REGULAR/ALLOMORPHIC/SUPPLETIVE/DEFECTIVE/PERIPHRASTIC. Never create a missing form just because a regular rule predicts one.
One-Sentence Answer
Translate suppletion and defective paradigms by preserving lexeme identity and grammatical function while allowing source irregularity to become target regularity—or source gaps to require target periphrasis.
Why This Precision Layer Matters
Irregular paradigms are dangerous precisely because fluent language models are good at analogy. A fabricated form can look completely plausible. Suppletive stems also invite lexical misanalysis because they do not resemble their lemma. Translation needs paradigm knowledge, not just string similarity.
What Suppletion Is
The first task is to identify the lexeme and the grammatical cell. Suppletion occurs when different cells of one lexeme’s paradigm are realised by forms that are not predictably related by ordinary phonology or morphology, as in English go/went or good/better. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating the suppletive form as an unrelated dictionary word and losing paradigm identity. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify whether the forms fill complementary grammatical cells of one lexeme. For example, went belongs to the past-tense paradigm of go despite its different stem. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Canonical Suppletion
This phenomenon belongs to paradigm structure rather than sentence meaning alone. Recent morphological work uses canonical typology to describe strong cases of suppletion where forms are maximally different yet unambiguously belong to one paradigm. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is assuming all irregularity is suppletion. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to compare formal predictability and paradigm function. For example, go/went is more strongly suppletive than a regular stem alternation caused by phonology. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Partial Suppletion
A precise translation should distinguish formal irregularity from functional meaning. Some alternations preserve part of the stem or show intermediate degrees of formal relatedness. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is forcing a binary regular/suppletive choice when the morphology lies on a continuum. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to compare stems across the whole paradigm. For example, a lexeme can show root replacement in some cells while sharing material in others. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Suppletion vs Allomorphy
The target may regularise a suppletive source or use periphrasis for a defective target. Allomorphy covers context-conditioned variant forms of a morpheme, while suppletion is an extreme type where alternants are highly different; theoretical boundaries vary. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is calling every conditioned allomorph suppletive. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to measure how predictable the alternation is from phonology/morphology. For example, English plural /s z ɪz/ allomorphy is not comparable to person/tense stem replacement. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Suppletion vs Stem Alternation
Morphological symmetry is not a translation requirement. Vowel change, consonant mutation or ablaut can create irregular stems without full suppletion. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating sing/sang/sung as the same kind of relation as go/went without qualification. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to ask whether a systematic phonological/morphological relation remains. For example, ablaut preserves recognisable stem identity even when irregular. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Lexical Paradigm Identity
The first task is to identify the lexeme and the grammatical cell. Translation must recognise that formally different source forms can belong to the same lexeme. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is creating separate target concepts because source forms look unrelated. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to map every surface form to its lexeme and paradigm cell. For example, two radically different source stems can both translate with forms of one English verb. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Tense Suppletion
This phenomenon belongs to paradigm structure rather than sentence meaning alone. Verbs frequently show suppletive stems according to tense/aspect categories. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is translating the suppletive past as a separate lexical event. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to record TENSE/ASPECT before lexical selection. For example, go/went differs in time, not basic motion lexeme. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Aspect-Based Suppletion
A precise translation should distinguish formal irregularity from functional meaning. Some languages use different stems depending on aspect rather than tense alone. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is assigning lexical semantic differences where the paradigm contrast is grammatical viewpoint. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify the aspectual cell. For example, perfective and imperfective stems can be suppletively related in a lexeme. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Person Suppletion
The target may regularise a suppletive source or use periphrasis for a defective target. Pronouns and certain verbs can use radically different forms across person categories. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is assuming formal difference means unrelated referents or lexemes. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to read person features from the paradigm. For example, first/second/third person sets often have historically unrelated stems but systematic paradigmatic roles. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Number Suppletion
Morphological symmetry is not a translation requirement. A lexeme can use different stems in singular, plural, dual or other number cells. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating plural suppletion as a distinct lexical meaning rather than number. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify the number feature before translating. For example, a motion verb can have one stem for singular subject and another for plural subject in some languages. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Degree Suppletion
The first task is to identify the lexeme and the grammatical cell. Adjectives can show suppletion across positive, comparative and superlative cells. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is translating better/best as unrelated evaluations rather than comparative/superlative of good. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify DEGREE feature. For example, good–better–best is one comparison paradigm. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Imperative Suppletion
This phenomenon belongs to paradigm structure rather than sentence meaning alone. Commands can use special stems that differ sharply from declarative forms. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is translating the imperative stem as a separate lexical command verb. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify imperative mood before dictionary lookup. For example, a suppletive command form can still belong to the same base lexeme. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Existential/Copular Suppletion
A precise translation should distinguish formal irregularity from functional meaning. Copulas and existential verbs often have unusually irregular paradigms, sometimes combining forms from historically distinct sources. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is forcing one root analysis through every cell. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to treat the modern paradigm synchronically. For example, English be/am/is/are/was/were shows extensive stem diversity. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Motion Verb Suppletion
The target may regularise a suppletive source or use periphrasis for a defective target. Go/come verbs are cross-linguistically frequent sites of suppletion, sometimes conditioned by tense, number, imperative or direction. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is assuming different motion stems necessarily encode different paths. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to check whether the grammar treats them as cells of one verb. For example, a plural motion stem can differ formally while retaining the same basic path meaning. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Defective Paradigms
Morphological symmetry is not a translation requirement. A defective paradigm lacks one or more forms that would otherwise be expected from the language’s inflectional system. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is inventing the missing form by analogy because the paradigm seems to demand it. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to mark the cell as unavailable or restricted. For example, English modal can lacks an ordinary infinitive *to can in its modal use and relies on periphrasis such as be able to. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Defectiveness vs Rare Forms
The first task is to identify the lexeme and the grammatical cell. A very rare, archaic or stylistically restricted form is not necessarily absent. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is declaring a cell defective simply because corpus frequency is low. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to check grammatical descriptions and attested usage. For example, a literary form can exist even if speakers prefer a periphrasis. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Defectiveness vs Lexical Gap
This phenomenon belongs to paradigm structure rather than sentence meaning alone. A lexical gap is absence of a word/concept label; paradigm defectiveness is absence of an inflectional cell for an existing lexeme. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is confusing ‘no word for X’ with ‘this verb has no imperative’. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to ask whether other cells of the same lexeme exist. For example, a verb can exist but lack a normal first-person form. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Defectiveness vs Blocking
A precise translation should distinguish formal irregularity from functional meaning. A regular-looking form can be blocked because another form or construction conventionally fills the function. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is generating the predictable form even when the grammar selects a competing expression. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify the established competitor. For example, a periphrastic form may occupy a function that morphology cannot or does not realise. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Periphrastic Repair
The target may regularise a suppletive source or use periphrasis for a defective target. Languages often fill defective cells with multiword constructions, another lexeme or a related paradigm. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating periphrasis as a semantic addition rather than the normal grammatical solution. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify what grammatical feature the periphrasis supplies. For example, be able to supplies infinitival/future possibilities where modal can is defective. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Auxiliary Defectiveness
Morphological symmetry is not a translation requirement. Modal and auxiliary verbs are common sites of restricted paradigms because they can lack infinitives, participles, imperatives or nonfinite forms. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is forcing an ordinary full-verb paradigm onto a grammaticalised auxiliary. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to list which nonfinite/finite forms are actually licensed. For example, English must/may/can have paradigm restrictions unlike lexical verbs. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Impersonal and Weather Verbs
The first task is to identify the lexeme and the grammatical cell. Some verbs are restricted to third-person or impersonal uses and therefore appear defective under a full personal paradigm. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is inventing first- or second-person forms in literal translation. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to check semantic and grammatical person restrictions. For example, weather verbs may lack normal personal paradigms except metaphorical uses. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Semantically Incompatible Cells
This phenomenon belongs to paradigm structure rather than sentence meaning alone. Certain inflectional combinations are absent because their meanings clash with the lexeme’s semantics or construction. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating the absence as arbitrary and creating an odd target form. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to ask whether the cell would make semantic sense. For example, some stative or impersonal lexemes resist imperative or agentive forms. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Morphological vs Semantic Defectiveness
A precise translation should distinguish formal irregularity from functional meaning. A cell can be morphologically missing even if its meaning can be expressed periphrastically; semantic expressibility therefore does not mean morphological completeness. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is assuming periphrastic availability proves the inflected form exists. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to separate morphological paradigm from functional expression. For example, a language can express the meaning but only through another construction. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Dialect Variation
The target may regularise a suppletive source or use periphrasis for a defective target. A form defective in one dialect or standard may exist in another variety. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is declaring a universal gap based on one reference grammar. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify target/source variety. For example, regional speakers can regularise or preserve forms differently. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Historical Change
Morphological symmetry is not a translation requirement. Suppletion and defectiveness can change over time through analogical levelling, grammaticalisation or paradigm replacement. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is using historical forms as if they were current or treating modern irregularity as timeless. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to date and variety-check unusual forms. For example, a once-productive form can become archaic while a periphrasis takes over. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Analogy and Regularisation
The first task is to identify the lexeme and the grammatical cell. Speakers can create regularised forms beside inherited suppletive or defective paradigms. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating learner/colloquial analogical forms as standard without register control. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to check normative and descriptive status. For example, regularisation can be productive in child language or dialect change. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Paradigm Competition
This phenomenon belongs to paradigm structure rather than sentence meaning alone. Two forms can compete for the same cell during change or dialect variation. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is assuming one must be an error. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify register, region and chronology. For example, an older suppletive form and newer regular form can coexist. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Suppletion in Translation into English
A precise translation should distinguish formal irregularity from functional meaning. A source suppletive paradigm may map onto a regular English lexeme, hiding the source irregularity completely. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is trying to reproduce formal irregularity instead of grammatical meaning. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to translate lexeme + features. For example, two source stems can both become forms of English go/come/etc.. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
English Suppletion into a Regular Target
The target may regularise a suppletive source or use periphrasis for a defective target. English go/went or good/better can map to target forms built regularly from one stem. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating the target’s regular morphology as loss of lexical identity. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to preserve tense/degree rather than source formal irregularity. For example, regular target past can faithfully translate suppletive English went. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Defective Source into Complete Target
Morphological symmetry is not a translation requirement. A source periphrasis required by a defective paradigm can map to a simple target inflection. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is overtranslating the source helper verb as extra lexical meaning. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify the periphrasis’s grammatical function. For example, source ‘be able to’ can map to a target verb with a normal infinitive/future. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Complete Source into Defective Target
The first task is to identify the lexeme and the grammatical cell. A target may lack the simple form the source uses, forcing periphrasis. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is inventing an unattested target inflection for symmetry. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to use the target’s established repair strategy. For example, source modal infinitive can require a lexical/periphrastic target equivalent. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Dictionaries and Lemmas
This phenomenon belongs to paradigm structure rather than sentence meaning alone. Dictionaries usually list suppletive forms under one lemma or cross-reference them, while defective forms may include paradigm warnings. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is looking up an unfamiliar suppletive form as a separate lemma and selecting the wrong sense. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to trace paradigm references and grammatical labels. For example, went redirects to go because it is a paradigm cell, not an independent modern lexeme. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Morphological Tagging
A precise translation should distinguish formal irregularity from functional meaning. NLP systems need to assign lemma and features correctly despite radically different surface forms or absent cells. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is treating a suppletive token as out-of-vocabulary or mislemmatising it. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to verify lemma + feature bundle. For example, went must be tagged as GO+PAST. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Machine Translation: Regularisation
The target may regularise a suppletive source or use periphrasis for a defective target. AI can invent a predictable regular form in a source/target language where the paradigm is defective or suppletive. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is accepting plausible morphology without attestation. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to check unusual cells against a reliable paradigm. For example, a model may fabricate an infinitive/imperative that native grammar avoids. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Machine Translation: Lexeme Split
Morphological symmetry is not a translation requirement. AI can treat suppletive forms as different lexemes and lose semantic continuity. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is translating one cell with a different concept. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to map the surface form to its lemma first. For example, a past suppletive stem must preserve the base verb’s lexical meaning. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Machine Translation: Periphrasis Literalism
The first task is to identify the lexeme and the grammatical cell. Models can translate every word of a periphrastic repair and make the target more complex than necessary. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is preserving grammatical scaffolding as lexical meaning. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to identify whether the target has a simple inflected equivalent. For example, the source helper can disappear when target morphology expresses the same cell. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Back-Translation Paradigm Check
This phenomenon belongs to paradigm structure rather than sentence meaning alone. A final audit should recover lexeme identity and grammatical cell rather than source stem shape. Paradigms are language-specific, so a form’s irregularity is evidence about morphology, not a semantic feature that must itself be recreated in the target.
The main translation risk is checking whether the same irregular stem relation appears. AI especially tends to complete paradigms by analogy, but native grammar can block or lack exactly the form that looks most predictable. Suppletive forms create the opposite problem: too little formal similarity can hide a shared lexeme.
A practical diagnostic is to write LEXEME + CELL + TARGET FORM TYPE. For example, translation is correct if the paradigm function survives, regardless of regularity. Once the lexeme/cell is secure, use the target’s ordinary paradigm—even if its regularity profile is completely different.
Worked Examples
Worked Example 1: Go / Went
Source problem: Source or target uses a suppletive past stem unrelated in form to the present lemma.
Diagnosis: Went is GO+PAST, not a separate motion lexeme.
Target strategy: Translate the lexeme and tense normally.
Verification: Do not choose another motion meaning simply because the stem differs.
Worked Example 2: Good / Better
Source problem: The comparative is suppletive.
Diagnosis: Better belongs to the comparison paradigm of good.
Target strategy: Use the target comparative of the equivalent adjective, regular or irregular.
Verification: Degree relation must survive; source irregularity need not.
Worked Example 3: Defective Modal
Source problem: A modal lacks a normal infinitive/nonfinite form and uses periphrasis.
Diagnosis: The function exists but the morphological cell is unavailable.
Target strategy: Use the source/target’s established periphrastic repair such as be able to.
Verification: Do not invent an unattested infinitive merely for symmetry.
Worked Example 4: AI Regularises a Gap
Source problem: A model fabricates a plausible first-person or imperative form that native speakers avoid.
Diagnosis: Analogy has filled a defective cell incorrectly.
Target strategy: Replace the fabricated form with an attested periphrasis or alternative lexeme.
Verification: Fluency is not evidence of paradigm legitimacy.
Practical Diagnostic Workflow
Create a PARADIGM MAP around the problematic token. Include lemma, tense/aspect/mood, person, number, degree or other relevant features. Add neighbouring cells. A suppletive relation becomes visible as one lexeme with radically different forms; defectiveness becomes visible as an unavailable cell surrounded by existing ones.
Then check the repair strategy. Does the language use another lexeme, auxiliary, periphrasis, stylistic avoidance or a restricted archaic form? Translate the function through the established strategy rather than generating morphology productively.
Verification Method
Back-read the target and identify the same lexeme and grammatical cell. If the target has a regular form where the source is suppletive, that is acceptable. If the target uses a periphrasis where a simple source form exists, verify that the target paradigm genuinely requires it. If a fabricated form appears, replace it.
Recovery Method
When a suppletive form was mislemmatised, return to the paradigm and restore lexical identity. When a defective form was invented, delete it and select the attested periphrastic or lexical repair. When periphrasis was translated word-for-word unnecessarily, reduce it to the target’s ordinary grammatical expression.
Quality-Assurance Checklist
- Identify lemma before translation.
- Identify exact paradigm cell.
- Distinguish suppletion from ordinary allomorphy.
- Distinguish suppletion from ablaut/stem alternation.
- Distinguish defectiveness from rarity.
- Distinguish defectiveness from lexical gaps.
- Check periphrastic repairs.
- Check auxiliaries and modals for restricted paradigms.
- Check dialect/register variation.
- Check historical vs current forms.
- Do not invent predictable forms.
- Do not split suppletive forms into new lexemes.
- Allow target regularisation.
- Allow target periphrasis.
- Audit AI paradigm completion.
Frequently Asked Questions
What is suppletion?
A paradigm relation where grammatical cells of one lexeme use forms that are not predictably related by ordinary morphology or phonology.
Is went a separate verb from go?
Synchronically in English it is the past-tense form in the paradigm of go.
Is every irregular verb suppletive?
No. Many irregular verbs use stem alternations or allomorphy that remain formally related.
What is a defective paradigm?
A paradigm in which one or more otherwise expected grammatical cells lack a normal form or are strongly restricted.
Does defectiveness mean the meaning cannot be expressed?
No. Speakers often use periphrasis or another lexeme to express the function.
Is syncretism the same as suppletion?
No. Syncretism uses one form for multiple cells; suppletion uses markedly different forms within one lexeme’s paradigm.
Can a defective form exist in one dialect but not another?
Yes. Defectiveness and regularisation can vary by dialect, register and historical stage.
Why does AI struggle with defective paradigms?
Generative models are excellent at analogy and can invent forms that look morphologically plausible but are unattested or unacceptable.
Where This Fits in the eduKate Translation Architecture
Route same-form/multiple-cell phenomena to Case Syncretism; general inflection/derivation to Word Formation and Morphological Meaning; number-conditioned stem replacement to the Grammatical Number owner; tense/aspect interpretation to Lexical Aspect and Telicity.
Final Principle
A paradigm is allowed to be irregular, incomplete or historically messy. Translation should not tidy it by invention. Identify the lexeme and cell, respect genuine gaps, and use the target language’s own regular, suppletive or periphrastic solution.