Cross-Situational Word Learning: Why Recognising a New Word Can Be Easier Than Producing It

A child hears a new word.

There are several objects nearby.

Nobody points.

The child does not know which object the word names.

Later, the same word appears again in a different scene.

Then again.

Across repeated encounters, one object keeps travelling with the word.

The learner begins to solve the mapping.

This is called cross-situational word learning.

It matters because real vocabulary learning is often:

messier than a flashcard.

Children and language learners frequently meet words before anyone gives them a clean definition. They gather evidence across moments.

A 2026 study of 95 children aged 5–9 examined what happens after that mapping is formed.

The central question was not only:

Can the child recognise the right object?

It was also:

Can the child produce the new word?

That distinction gives this article one clear job:

to explain why receptive word learning can look successful before the new word is strong enough for accurate spoken production.

Quick answer

Recognition and production are:

different tests of vocabulary strength.

A learner may know enough to choose:

the correct referent

without knowing enough to:

say the new word accurately.

The 2026 study by Kaityn Contino, Anna Duncan and colleagues found that monolingual and bilingual children performed similarly on receptive cross-situational word learning, but expressive production revealed clearer differences.

Within bilingual children, longer bilingual experience predicted:

stronger receptive

and:

stronger expressive performance.

Speaker variability did not significantly change outcomes in this particular child sample.

The lesson is:

do not treat one successful recognition test as complete word knowledge.

What is cross-situational word learning?

Imagine three unfamiliar objects:

A, B and C.

The learner hears:

dax.

Which object is dax?

Unknown.

Next trial:

B, C and D appear.

Again:

dax.

Now:

B and C remain possible.

Next trial:

B, E and F.

dax.

Across situations:

B is the consistent referent.

The learner can infer:

dax → B.

This is statistical learning across encounters.

The world rarely labels everything cleanly

A toddler hears:

“Put the ladle beside the pot.”

They may know:

pot.

They may not know:

ladle.

The room contains:

• a spoon;
• a bowl;
• a cloth;
• a ladle.

No dictionary appears.

The learner uses:

• repetition;
• co-occurrence;
• attention;
• context;
• later encounters.

Vocabulary grows from:

accumulated evidence.

Recognition is easier because the answer is present

Suppose the learner is asked:

Which one is the dax?

They see:

four objects.

The correct referent is:

on the screen.

The learner can use:

partial knowledge.

Now remove all objects and ask:

What was this object called?

The learner must produce:

dax.

Production requires:

• retrieving the sound form;
• recovering sequence;
• assembling pronunciation;
• doing so without visible alternatives.

This is a stronger demand.

The 2026 study separated receptive and expressive learning

Contino and colleagues studied:

95 children aged 5–9 years.

The children completed a cross-situational word-learning task under:

• single-speaker;
• multiple-speaker conditions.

The researchers compared:

• monolingual children;
• bilingual children.

They also looked within bilingual learners at:

length of bilingual experience.

This is important because:

bilingualism is not one identical state.

Receptive performance was broadly comparable

On receptive learning:

monolingual and bilingual children showed:

comparable performance.

Speaker variability did not significantly alter receptive accuracy.

Within the bilingual group:

longer bilingual experience predicted:

stronger receptive learning.

This tells us that:

a broad group label may hide:

meaningful individual experience.

Production revealed a different pattern

On expressive production:

monolingual children produced the novel words more accurately than bilingual children overall.

But within bilingual children:

longer bilingual experience was associated with:

higher expressive accuracy.

That is subtle.

It means:

“bilingual” should not be treated as a simple deficit or advantage label.

Experience matters.

Partial production still reveals learning

A child may hear:

mipanu.

Later they produce:

mitanu.

Incorrect?

Yes.

No learning?

No.

The production may reveal:

a partially established phonological representation.

This is why vocabulary assessment should distinguish:

nothing known

from:

partially known.

A word can exist at several strengths

Target:

meticulous.

A learner may:

1. recognise it as familiar;
2. choose the correct definition;
3. understand it in a sentence;
4. pronounce it approximately;
5. retrieve it from meaning;
6. use it appropriately in writing;
7. use it spontaneously in speech.

These are:

different lexical achievements.

Vocabulary is not:

known / unknown.

It is:

layered accessibility.

Why bilingual experience can matter without creating one universal bilingual effect

Bilingual learners coordinate:

more than one language system.

But bilingual experience varies in:

• age of exposure;
• amount of exposure;
• dominant language;
• contexts of use;
• school language;
• home language;
• speaking frequency.

So the question:

“Are bilingual children better at word learning?”

is often too crude.

A better question is:

Which word-learning outcome, under which input conditions, for learners with what language experience?

Singapore relevance

This matters naturally in Singapore.

A child may encounter:

English at school, Mandarin at home, Malay with relatives, Tamil with grandparents, or another family language in daily life.

Different children may be:

equally “bilingual” on a form

but:

very different in actual language experience.

Teaching should therefore observe:

what the child can recognise

and:

what the child can produce.

Speaker variability: do different voices help?

Vocabulary appears through:

different voices.

One teacher says:

photosynthesis.

A video narrator says:

photosynthesis.

A classmate says:

photosynthesis.

Acoustic details differ.

The learner must discover:

the same word beneath variable speech.

The 2026 result: no significant speaker-variability effect here

In this particular study:

single versus multiple speakers did not significantly change receptive or expressive outcomes.

This is useful because:

earlier studies have not always produced the same pattern.

So we should not state:

“multiple speakers always improve word learning.”

The evidence is:

context-sensitive.

One study should not become a universal rule

The participants were:

children aged 5–9.

The task involved:

novel word–referent mappings.

That is not identical to:

• Secondary vocabulary;
• academic words;
• essay vocabulary;
• subject terminology;
• adult second-language learning.

The study reveals:

a mechanism boundary.

Not:

a complete curriculum.

Recognition can flatter vocabulary estimates

Imagine a multiple-choice test:

concur means:

• agree;
• destroy;
• measure;
• hide.

The learner selects:

agree.

Now ask:

What verb means:

to agree with an opinion?

Silence.

Recognition is real knowledge.

But it can overstate:

production readiness.

Production can also underestimate understanding

The reverse mistake is possible.

A learner cannot say:

conscientious.

But can identify:

careful, responsible, morally serious behaviour

and understand the word in reading.

Production failure does not mean:

zero semantic knowledge.

Assessment should locate:

the weak layer.

Primary English

Target:

fragile.

First:

show several objects and ask:

Which one is fragile?

Later:

show the glass ornament and ask:

What word describes something that breaks easily?

Recognition → production.

Secondary English

Target:

reluctant.

Recognition:

Which sentence shows reluctance?

Production:

Complete:

“She was ______ to speak because…”

Independent use:

write one new situation.

General Paper

Target:

disproportionate.

Recognition:

identify the meaning in a policy paragraph.

Production:

use it accurately in:

“The burden falls disproportionately on…”

Now the learner must control:

word family

and:

grammar.

Science

Target:

permeable.

A student may recognise:

a permeable membrane

but fail to explain:

why a material is described as permeable.

Subject vocabulary needs:

recognition + explanation + use.

Mathematics

Target:

equivalent.

Recognition:

1/2 and 2/4 are equivalent.

Production:

explain:

what “equivalent” means.

Transfer:

use the word correctly for:

fractions, expressions or ratios.

Humanities

Target:

legitimacy.

Recognition:

identify which passage refers to accepted right to rule.

Production:

explain:

how legitimacy differs from:

power.

This moves from:

word recognition

to:

conceptual control.

Diagnosis before prescription

Student recognises words but rarely uses them

Diagnosis: receptive knowledge exceeds productive accessibility.

Repair: add meaning-to-form retrieval and sentence production.

Student produces approximate forms

Diagnosis: lexical representation may be partial rather than absent.

Repair: compare sound sequence, spelling and meaning; then retrieve again later.

Student performs poorly on a production test

Diagnosis: the result may reflect retrieval or phonological difficulty, not zero comprehension.

Repair: test recognition separately before deciding what is missing.

Student gets recognition items correct by guessing

Diagnosis: multiple-choice support may be carrying performance.

Repair: use delayed free recall and new-context interpretation.

Parent compares bilingual and monolingual children as fixed groups

Diagnosis: language experience has been flattened into a label.

Repair: consider actual exposure, proficiency and use across languages.

Teacher assumes multiple speakers always improve learning

Diagnosis: a plausible generalisation exceeds the evidence.

Repair: use varied voices for authentic listening, but verify learning outcomes rather than assuming benefit.

A practical recognition-to-production ladder

Target:

inevitable.

1. Recognition

Which means:

cannot be avoided?

2. Context identification

Which sentence uses it correctly?

3. Cued production

Some degree of change was ______.

4. Definition-to-word retrieval

What adjective means:

certain to happen and impossible to avoid?

5. Independent sentence

Use it in:

History.

6. Contrast

Explain:

inevitable

versus:

likely.

Now vocabulary has moved from:

recognition

toward:

usable precision.

Parents: ask two questions

When a child says:

“I know that word,”

ask:

1. Can you recognise it in a sentence?
2. Can you use or explain it without the answer in front of you?

Those questions reveal:

different layers.

Teachers: measure vocabulary with more than one response format

A strong vocabulary check may include:

• recognition;
• meaning recall;
• form recall;
• pronunciation;
• sentence use;
• new-context interpretation.

No single format captures:

the whole lexical representation.

AI-assisted vocabulary practice

A useful public prompt is:

“Teach me eight new words in stages. First test recognition from context. Then remove the options and test meaning recall. Then show the concept and make me retrieve the word form. Then ask me to say or type the word without a model. Finally, place it in a new subject context. Keep separate records of what I can recognise and what I can independently produce.”

This makes AI distinguish:

performance type.

A quiet literary lens

A high-level Hilary Mantel lens is useful because recognition often arrives before speech.

You know:

the face.

You know:

the office.

You know:

the significance.

But the name does not come.

Vocabulary can inhabit that same distance:

between:

I know it

and:

I can call it back.

Internal-link opportunities

• Bilingualism and English Vocabulary Learning
• Aural Lexical Knowledge
• Lexical Quality
• Tip of the Tongue and Lexical Retrieval
• How Language Works

Connections eduKateAI can learn

Cross-situational learning ↔ accumulated evidence: learners can build word–referent mappings by tracking recurring associations across ambiguous encounters.

Recognition ↔ partial lexical knowledge: a learner can identify the correct referent before the word form is strong enough for independent production.

Production ↔ lexical accessibility: speaking or writing a new word requires stronger retrieval of form than recognition tasks require.

Bilingualism ↔ experience gradient: categorical labels can hide meaningful variation in amount and duration of dual-language experience.

Speaker variability ↔ input variation: different voices alter acoustic form, but their learning effect is not universally positive or negative across studies.

Partial production ↔ representation strength: an approximate word form can reveal emerging phonological knowledge rather than complete absence of learning.

Subjects ↔ productive vocabulary: subject learning improves when learners can move from recognising terminology to explaining and using it independently.

AI language learning ↔ multi-format assessment: systems should track recognition, recall and production separately instead of storing one binary “knows word” state.

Final checkpoint

Can a child know a new word without being able to say it accurately?

Yes.

The 2026 study shows why:

receptive and expressive learning can:

diverge.

A useful rule is:

recognition tells us that a mapping exists; production tells us more about how independently accessible the word form has become.

Good vocabulary teaching therefore moves through:

recognise → retrieve → produce → transfer.

Research and reference basis

• Contino, K., Duncan, A. V., Weaver, H., Khandelwal, M., Boucher, A., & Crespo, K. (2026). From Recognition to Production: Receptive and Expressive Cross-Situational Word Learning in Monolingual and Bilingual Children. Behavioral Sciences, 16(7), 1080. Published 1 July 2026. https://doi.org/10.3390/bs16071080
• PubMed record: https://pubmed.ncbi.nlm.nih.gov/42510201/
• Crespo, K., & Kaushanskaya, M. (2025). Earlier research on speaker variability and cross-situational word learning provides an important comparison showing that effects of variability can differ by age and design.

This article deliberately owns the recognition-to-production gap in cross-situational novel-word learning, including how bilingual experience should be interpreted. It does not replace eduKateSG’s broader bilingual-vocabulary, listening or lexical-retrieval pages.