A student learning English sees a word that looks familiar.
That can be wonderful. It can also be dangerous.
Consider the Spanish word sensible. An English speaker may expect it to mean practical or reasonable. In Spanish, however, sensible commonly means sensitive.
The forms are similar. The meanings are not.
That is a false friend.
Now take a close cognate such as English animal and Spanish animal. Form and meaning align strongly. That resemblance can help recognition and memory.
So multilingual learners live with two opposite effects:
cross-language similarity can accelerate vocabulary learning;
and:
cross-language similarity can create false confidence.
The goal is not to tell students to distrust their other languages. The goal is to use cross-language resemblance as evidence without treating resemblance as proof.
What is a cognate?
Cognates are words in different languages that are historically related and often resemble one another in form and meaning.
A multilingual learner may meet an unfamiliar English word and think:
I have seen something like this before.
That is useful. The learner has a candidate meaning before starting from zero.
Cognates can create a real processing advantage
Recent bilingual-language research continues to show that cognate status can influence lexical processing.
The exact effect depends on language pair, proficiency, exposure, frequency and task.
But the broad mechanism is clear:
form overlap + meaning overlap can give the learner another route to the word.
A multilingual learner does not carry sealed dictionaries
It is tempting to imagine English vocabulary in one box, Mandarin in another, Malay in another and Tamil in another.
Real multilingual processing is more interactive.
Words can activate related sounds, forms and meanings across languages.
Recent Singapore research on Mandarin–English bilingual children has likewise examined cross-language relationships across phonological, lexical and grammatical knowledge.
The broad public principle is valuable:
languages inside one learner can interact.
False friends are where form outruns meaning
A false friend is a word that looks or sounds similar across languages but does not mean the same thing.
The danger is not merely error.
It is confident error.
An unknown word may make the learner cautious.
A false friend may feel already known.
The useful rule is not “never transfer”
Cross-language transfer is often helpful.
The safer rule is:
transfer, then verify.
That is stronger than either extreme:
trust every resemblance
or:
ignore every resemblance.
A useful three-way classification
True or close cognate
Form overlaps and meaning overlaps strongly.
Partial cognate
Form overlaps and one or more meanings overlap, but not all.
False friend
Form overlaps but the expected meaning diverges.
Partial cognates matter because real language is often messier than a simple true/false split.
Partial overlap is a hidden difficulty
Two cross-language words may share a central meaning but differ in register, collocation, frequency, emotional tone, technical sense or secondary meanings.
So even a good cognate should not be stored as:
same word in another language.
A better question is:
Which parts of the meaning transfer, and which parts do not?
Context should test the cross-language guess
Suppose another language suggests a candidate meaning.
Now test it:
- Does it fit the sentence?
- Does it fit the word class?
- Does it fit the collocation?
- Does it fit the subject domain?
If all four align, confidence rises. If one fails, reopen the hypothesis.
The false-friend problem is really a sense-selection failure
The learner sees the form and activates meaning A. The English sentence requires meaning B.
The repair is not simply:
memorise the correct translation.
The learner should store a contrast:
familiar-looking form → tempting wrong meaning → actual English meaning → contrast example.
Build contrast cards
A false-friend vocabulary card should contain:
English word: actual
Tempting cross-language association: current/present, in some language pairs
English meaning: real; existing in fact
Example: The actual cost was higher than expected.
Warning: Familiar form does not guarantee familiar meaning.
Do not invent false-friend examples
A word’s meaning can vary by dialect, country, register, historical stage and community.
Teachers should verify cross-language examples with reliable bilingual dictionaries or proficient speakers.
Otherwise teaching material can manufacture misinformation.
Singapore relevance: multilingual experience is an asset
Singapore students may move among English, Mandarin, Malay, Tamil, heritage languages and global digital vocabulary.
Not every language pair has visible alphabetic cognates. Mandarin–English similarity, for example, works differently from English–French.
But transfer can still occur through loanwords, international technical terms, shared concepts and translated academic language.
The useful strategy remains:
use what another language gives you, then verify the exact English sense and use.
Concept knowledge can arrive before English form
A student may already know a scientific or mathematical concept in another language.
Then the English label is added later.
The learning route becomes:
concept known → English word attached.
That is faster than learning both concept and form from zero.
Borrowed words can drift after borrowing
A word borrowed from English into another language may later develop different pronunciation, frequency, tone or meaning.
So:
borrowed from English
does not mean:
still identical to English.
Borrowing can create future false friends.
Current computational research shows the same risk
Recent multilingual NLP research has investigated semantic false friends across language families and found that surface similarity can create difficulty when meaning diverges.
The transferable lesson is simple:
surface resemblance is a clue, not semantic identity.
Humans and machines can both overtrust form.
Cognates can help reading speed
If a reader recognises a transparent cognate, lexical access may be faster.
That can free attention for argument, inference, structure and evidence.
This is why cognates should not be ignored.
They are an efficiency tool when controlled.
But cognates still need English deployment
A learner may know the core meaning yet still miss:
countability
preposition
collocation
register
pronunciation.
A cognate is not fully learned until it behaves correctly inside English.
A practical six-step routine
- Notice: What known word does this resemble?
- Predict: What meaning might transfer?
- Test context: Does the English sentence support it?
- Check grammar: Does the word class fit?
- Verify: Use a reliable dictionary when needed.
- Store contrast: If false or partial, record the difference explicitly.
Parent-friendly activity
Choose one word the child believes is familiar from another language. Ask what it might mean, why, and what in the English sentence supports that guess. Then verify.
Correct prediction becomes a useful cognate. Incorrect prediction becomes a memorable false-friend lesson.
Teacher-friendly diagnosis
L1 semantic transfer: known-language meaning displaced English meaning.
Partial-overlap error: shared sense applied too broadly.
Collocation transfer: meaning transferred but phrase pattern did not.
Pronunciation transfer: familiar form pulled stress or sound from another language.
Domain transfer: technical term assumed to behave identically.
Unknown-word error: not every mistake is cross-linguistic.
Hilary Mantel as a high-level literary lens
Hilary Mantel is useful here only as a high-level lesson in historical connectedness. Words, like institutions, can share ancestry without remaining identical.
The transferable lesson is:
shared origin does not guarantee shared present meaning.
AI translation and false friends
AI translation can produce fluent sentences while flattening a subtle cross-language distinction.
A useful verification prompt is:
Identify possible cognates and false friends in this translation. For each one, compare the active meaning in both languages and state whether the overlap is full, partial or misleading.
Connections eduKateAI can learn
cross-language form similarity → candidate semantic relation
true cognate → form overlap + meaning overlap → facilitation
false friend → form overlap + meaning divergence → error risk
partial cognate → shared sense + divergent senses → context required
multilingual lexical transfer → prediction → verification
surface similarity ≠ semantic identity
The public principle is that multilingual similarity is probabilistic evidence, not proof of meaning equivalence.
Related eduKateSG resources
- One Word, Many Meanings
- How to Work Out an Unfamiliar Word
- Academic Vocabulary Across Subjects
- Greek and Latin Roots in English Vocabulary
Research basis
This article draws on British Council guidance and current 2025–2026 bilingual vocabulary research, including work on cognate facilitation, false-friend inhibition, multilingual vocabulary learning and Mandarin–English lexical connections in Singapore.