Suppletion in translation is the extreme end of irregular morphology: one lexeme uses two or more formally unrelated or highly divergent stems inside the same grammatical paradigm. English go/went, be/am/is/was, and good/better/best are familiar examples. The grammatical relationship is regular—past of GO, comparative of GOOD—even though the forms themselves do not transparently derive from one another.
People searching for suppletion, suppletive forms, go went suppletion, good better best morphology, irregular paradigm, stem allomorphy are solving a lexical-identity problem. A translation system that treats every unfamiliar stem as a new dictionary entry can split one lexeme into several unrelated meanings. The opposite error is to regularize the source mentally and overlook distinctions in person, number, tense, comparison, ordinal number or honorific status that the suppletive forms actually encode.
This article owns the suppletion and extreme-allomorphy precision layer for eduKateSG. It does not replace Word Formation, Case Syncretism or ordinary inflection. Its narrow task is to preserve lexical identity across formally discontinuous paradigms, distinguish true suppletion from predictable phonological alternation, recover grammatical features from stem choice, and prevent irregular source forms from becoming unrelated target lexemes.
Quick Read
Build PARADIGM first, translate second. Put all forms that belong to one lexeme in one table with TENSE, PERSON, NUMBER, DEGREE, CASE or REGISTER. If the form changes radically while grammatical meaning remains within one paradigm, preserve lexical identity.
One-Sentence Answer
Translate suppletion by recognizing that radically different source forms can be grammatical members of one lexeme rather than separate vocabulary items.
Why This Precision Layer Matters
Suppletion defeats form-based translation. The strongest clue to lexical identity—the shared stem—can disappear entirely. Without paradigm awareness, GO and WENT look unrelated, GOOD and BETTER look unrelated, and forms of BE look like several verbs. Translation therefore has to trust grammatical distribution and paradigm structure more than surface resemblance.
What Suppletion Is
Suppletion occurs when a lexeme uses unexpectedly different forms for grammatical cells that would normally be related by inflection.
The main translation risk is creating separate lexical entries for each stem. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to group forms by paradigm and grammatical function. For example, GO and WENT belong to one lexeme despite different stems. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate lexical identity first and grammatical cell second. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Canonical Suppletion
Canonical accounts place the clearest cases at the extreme of formal difference combined with strong paradigmatic unity.
The main translation risk is forcing every irregular alternation into one yes/no definition. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to place examples on a continuum from regular allomorphy to extreme suppletion. For example, GO/WENT is more extreme than a predictable vowel alternation. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve the grammar even when theoretical classification is borderline. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Suppletion vs Allomorphy
Allomorphy covers multiple forms of the same morpheme or stem; suppletion is commonly treated as its most extreme and unpredictable end.
The main translation risk is calling every irregularity suppletion. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to ask whether alternation is phonologically predictable or lexically conditioned. For example, Leaf/leaves differs from go/went in degree of formal predictability. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate the paradigm even if the label remains theoretical. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Go and Went
English GO uses WENT for its ordinary past-tense cell.
The main translation risk is translating WENT as a separate lexical verb. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to label tense before dictionary lookup. For example, Went in “she went home” is past GO. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, use the target past of GO, whatever its target form. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
The BE Paradigm
English BE displays unusually extensive suppletion across person, number and tense.
The main translation risk is treating AM, IS, ARE, WAS and WERE as unrelated copulas. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to map each cell to one BE lexeme. For example, I am, you are and he is differ formally but share one lexical core. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, generate target copular forms from the grammatical features. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Good Better Best
Comparative and superlative suppletion can replace ordinary degree morphology.
The main translation risk is treating BETTER as semantically unrelated to GOOD. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to write POSITIVE, COMPARATIVE and SUPERLATIVE cells. For example, Better is the comparative of good in its ordinary evaluative sense. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, use the target comparative system. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Bad Worse Worst
Negative-evaluation adjectives can also use suppletive degree forms.
The main translation risk is losing the comparative relation. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to map the scale and degree explicitly. For example, Worse means more bad in the relevant comparison. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve the comparative direction, not just negative evaluation. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Person Suppletion
Pronouns and some predicates can use radically different forms across person values.
The main translation risk is mistaking person change for lexical change. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to record PERSON before interpreting stem identity. For example, First-, second- and third-person forms can share a paradigm with little formal overlap. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate the correct speech-act role. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Number Suppletion
A lexeme can select different stems across singular, plural, dual or other number cells.
The main translation risk is treating a number form as a separate lexeme. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to map the full number paradigm. For example, A language can use unrelated singular and plural stems for the same noun or verb. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve count and lexical identity together. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Tense Suppletion
Different tense cells can select unrelated stems.
The main translation risk is looking for an affix that does not exist. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to identify tense from the paradigm, syntax and context. For example, GO/WENT shows past by root replacement rather than a suffix. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate the source tense even if target morphology is regular. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Aspect Suppletion
Perfective, imperfective or other aspectual categories can trigger different stems in some languages.
The main translation risk is assuming the new stem denotes a different lexical event. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to compare aspect cells within one lemma. For example, One verb can use distinct stems for completed and ongoing events. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, keep event identity stable while preserving aspect. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Mood Suppletion
Imperative, subjunctive or irrealis cells can be suppletive.
The main translation risk is translating a special mood stem as a different command verb. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to identify clause mood first. For example, A dedicated imperative stem can still mean the same underlying action. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve directive or irrealis force. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Comparative Suppletion
Adjectival degree is a major cross-linguistic domain of suppletion.
The main translation risk is regularizing away the source degree contrast. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to build a degree paradigm. For example, Good/better/best is one scale expressed by divergent forms. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate positive, comparative and superlative correctly. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Ordinal Suppletion
Cardinal-to-ordinal relations can be suppletive, as in one/first and two/second.
The main translation risk is treating ordinal forms as unrelated number words. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to pair cardinal and ordinal functions. For example, First is the ordinal counterpart of one. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve ranking and numeral identity. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Case Suppletion
Pronouns and nominals can have highly divergent forms across case.
The main translation risk is treating subject and object forms as different referents. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to track person, number and case together. For example, I/me/my refer to one participant under different grammatical relations. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, use target case or syntactic forms for the same referent. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Honorific Suppletion
Some languages replace ordinary verbs with distinct honorific or humble lexemes in socially marked paradigms.
The main translation risk is turning honorific replacement into a different event. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to record social register alongside lexical meaning. For example, A respectful verb can replace the ordinary verb SAY, COME or EAT. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve the core event and social stance. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Historical Origin
Suppletive forms often come from historically different lexemes that merged into one paradigm.
The main translation risk is reviving obsolete historical meanings in ordinary translation. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to separate diachronic origin from synchronic grammar. For example, WENT historically belongs to WEND but synchronically serves GO. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate modern paradigm function unless the source is etymological. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Synchronic Paradigm
For translation, current speakers’ grammatical system matters more than historical composition.
The main translation risk is letting etymology override present lexical identity. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to ask what lemma a modern grammar or dictionary assigns. For example, Historically unrelated forms can function as one modern lexeme. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate the synchronically active paradigm. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Morphological Gaps
Suppletive forms often fill cells where a regular form is absent.
The main translation risk is hallucinating a regular inflected form. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to check the attested paradigm. For example, Standard English uses went rather than goed. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, use real grammatical forms. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Dialect Variation
Suppletion and irregular paradigms can vary by dialect, register and historical period.
The main translation risk is correcting a legitimate variety before interpreting it. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to identify source variety and date. For example, Regional past forms may differ from standard ones. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve meaning and appropriate register. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Lexical Ambiguity
A suppletive form can resemble or historically derive from another lexeme.
The main translation risk is selecting the wrong dictionary sense because of surface form. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to use syntax and paradigm position jointly. For example, Modern went normally belongs to GO despite historical WEND. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, resolve current grammatical function before lexical translation. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Dictionary Cross-References
Dictionaries commonly list suppletive forms as cross-references to a lemma.
The main translation risk is failing to follow the cross-reference. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to inspect morphological metadata and irregular-form labels. For example, Better may point the reader to GOOD. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate from lemma plus grammatical features. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Frequency and Entrenchment
High-frequency words often preserve highly irregular forms because speakers learn them as entrenched paradigm cells.
The main translation risk is treating high irregularity as semantic exceptionalism. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to separate frequency from meaning. For example, BE remains extremely irregular while retaining ordinary copular functions. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate the semantics normally. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Language Learning Contexts
Suppletive sets are often memorized as lists but become clearer as paradigms.
The main translation risk is teaching forms as unrelated vocabulary. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to organize examples by cells and contrasts. For example, Go/went and good/better become structurally intelligible when placed in paradigms. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, preserve explanatory relationships in educational translation. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Translation into a Regular Target
A suppletive source paradigm may map to a regular target paradigm.
The main translation risk is trying to reproduce irregularity for its own sake. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to translate lemma and grammatical cell. For example, A source suppletive past can map to target root plus regular past suffix. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, allow target regularity. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Translation into a Suppletive Target
The target may itself use suppletion but in different paradigm cells.
The main translation risk is mapping source stem A directly to target stem A. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to recompute the target form from features. For example, Source tense suppletion may map to target person suppletion. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, generate target morphology independently. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Metalinguistic Translation
When the source explicitly discusses irregularity, the formal contrast itself becomes part of the content.
The main translation risk is normalizing away the very phenomenon under discussion. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to preserve forms and add explanation as needed. For example, A textbook comparing go and went needs both forms visible. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, retain source examples and translate the explanation around them. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
AI Lexeme-Splitting Bias
Models can assign different meanings to formally unrelated paradigm members.
The main translation risk is creating inconsistent translations for one lemma. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to normalize every form to a lemma before generation. For example, Went and go should resolve to GO plus tense. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, translate consistently across the paradigm. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
AI Over-Regularization
Models can hallucinate regular forms where the language requires suppletion.
The main translation risk is producing unattested morphology. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to verify real paradigm cells. For example, Goed is not the standard past of go. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, replace generated regular forms with attested target forms. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
AI Etymology Bias
Models with lexical knowledge can overemphasize historical origin.
The main translation risk is translating a modern suppletive form according to its source lexeme. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to prioritize synchrony. For example, WENT should not be translated as WEND in ordinary modern English interpretation. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, use current grammar. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Quality Control Across Documents
Repeated occurrences of different suppletive forms should remain terminologically coherent.
The main translation risk is translating one paradigm member inconsistently across a document. This kind of error survives ordinary fluency review because the target may still read smoothly while the underlying grammar has changed. A precision translation therefore checks the hidden relation before polishing style.
A practical diagnostic is to maintain a lemma-level terminology record. For example, Present and past forms can appear far apart while belonging to one key concept. Once that distinction is secure, the target can express it with ordinary words, inflection, auxiliaries, adverbs, word order, periphrasis, or no overt counterpart at all.
For recovery, verify consistency at lemma level, not surface-form level. The target should preserve the source-supported distinction without inventing extra participants, events, lexical identities, or grammatical features.
Worked Examples
Worked Example 1: Go/Went
Source problem: A source verb uses unrelated present and past stems.
Diagnosis: Both occupy cells of one lexeme.
Target strategy: Translate both with the target equivalent of GO, inflected appropriately.
Verification: Do not create two lexical verbs.
Worked Example 2: Good/Better
Source problem: A comparative stem is unrelated in shape to the positive adjective.
Diagnosis: The form is comparative degree on the same scale.
Target strategy: Use the target comparative of GOOD.
Verification: Verify that comparison is explicit.
Worked Example 3: Suppletive Pronoun Case
Source problem: Subject and object pronoun forms share little formal material.
Diagnosis: They refer to the same discourse participant under different case roles.
Target strategy: Use the target pronoun form required by syntax.
Verification: Do not create two referents.
Worked Example 4: Honorific Replacement
Source problem: A respectful context selects a different lexical verb for the same event.
Diagnosis: The event meaning remains stable while social stance changes.
Target strategy: Use a respectful target equivalent or register strategy.
Verification: Do not turn politeness into a different event.
Practical Diagnostic Workflow
Build a SUPPLETION PARADIGM with LEMMA, SURFACE FORM, PERSON, NUMBER, TENSE, ASPECT, MOOD, DEGREE, CASE and REGISTER. Do not translate an isolated irregular form before locating its paradigm cell.
Run the regular-form counterfactual test. Ask what a regular inflected form would look like and whether the language instead requires a different stem. If so, treat the unexpected form as paradigm evidence rather than as a new lexeme.
Verification Method
From the target, identify whether all suppletive source forms were translated as members of the same lexical paradigm where appropriate. Check tense, degree, person, number, case and honorificity separately.
Recovery Method
If one source lexeme was split into several target meanings, normalize all forms to the source lemma and regenerate by grammatical features. If the target invented regular forms, replace them with attested target paradigm cells.
Practice Sequence
Build mini-paradigms for GO, BE, GOOD and a hypothetical pronoun with unrelated case forms. Translate them into a target with regular morphology, then reverse the exercise with a regular source and suppletive target.
Quality-Assurance Checklist
- Identify the lemma.
- Locate the paradigm cell.
- Distinguish suppletion from predictable phonology.
- Track tense.
- Track person and number.
- Track comparison degree.
- Track case.
- Track honorific register.
- Separate historical origin from synchronic function.
- Respect dialect variation.
- Use attested forms.
- Normalize lexeme identity before target generation.
- Audit AI lexeme splitting.
- Audit AI over-regularization.
- Preserve irregularity in metalinguistic translation.
Frequently Asked Questions
What is suppletion?
The use of radically different forms within the grammatical paradigm of one lexeme.
Why is go/went suppletive?
Went functions as the standard past-tense form of go even though the two forms are not transparently related in shape.
Is all irregularity suppletion?
No. Predictable or less extreme allomorphy can be irregular without being canonical suppletion.
Is good/better suppletion?
Yes, it is a classic example of comparative suppletion.
Can pronouns be suppletive?
Pronoun paradigms can contain highly divergent person, number or case forms.
Does etymology determine translation?
Usually no. Modern paradigmatic function takes priority unless the text is historical or metalinguistic.
Should target-language irregularity match source irregularity?
No. Translate grammatical meaning and then generate the target’s own paradigm.
Why does AI fail?
Surface-form models can miss lexical unity when paradigm members share little or no orthographic material.
Internal Routes Through the eduKate Translation Architecture
Route general morphology to Word Formation and Morphological Meaning: https://edukatesg.com/2026/09/18/translate-precisely-word-formation-morphological-meaning-prefixes-suffixes-inflection-derivation/. Route identical-form ambiguity to Case Syncretism and Morphological Ambiguity: https://edukatesg.com/2026/09/19/translate-precisely-case-syncretism-morphological-ambiguity-one-form-multiple-cases/. Route person/number agreement to Grammatical Gender and Agreement: https://edukatesg.com/2026/09/18/translate-grammatical-gender-agreement-reference-gender-number-person-differently/.
Final Principle
Suppletion breaks formal continuity without breaking lexical identity. Translate the paradigm, not the stem: recover the lemma and grammatical cell first, then let the target language express that same cell through its own regular or irregular system.