Look at:
farmer.
English learners can analyse it:
farm + -er.
The meaning is transparent:
a person associated with farming.
Now look at:
corner.
It contains the letters:
corn + -er.
But a corner is not:
a person who corns.
Now:
deliver.
It appears to contain:
de- + liver.
But its modern meaning is not composed from:
a prefix de-
plus:
the organ liver.
These are examples of morphological illusions: forms that tempt the reader to segment a word into familiar pieces even when those pieces are not the word’s actual morphological structure.
The central principle is:
Visible or audible resemblance to a known morpheme is evidence for a possible analysis, not proof that the word is genuinely built from that morpheme.
What is a morphological illusion?
A morphological illusion occurs when a word contains a sequence that looks like:
root + affix
or:
prefix + base
but the apparent decomposition does not correctly explain the word’s meaning or history.
Examples often discussed in psycholinguistics include:
corner
brother
deliver
depending on the specific experimental definition.
These forms are sometimes called:
Why do readers see morphemes that are not really there?
The reading system is efficient.
When it sees a familiar-looking ending such as:
-er,
it can rapidly test:
Is this a stem plus a suffix?
That heuristic is useful for:
teacher
farmer
runner
builder.
But the same heuristic can misfire on:
corner.
The brain often decomposes before full meaning arrives
Psycholinguistic research has repeatedly found evidence that readers can become sensitive to:
possible morphological structure
very early during word recognition.
This is useful because many English words are genuinely complex.
But fast decomposition is:
a hypothesis generator.
Semantic interpretation still has to confirm or reject the analysis.
“Corner” is orthographically deceptive
The letters:
corn
and:
-er
are both familiar.
Yet:
corner
is not synchronically built from:
corn + agentive -er.
The apparent suffix boundary is:
accidental from the perspective of modern morphology.
“Deliver” shows why prefix spotting can fail too
The form:
deliver
may tempt a learner to identify:
de-.
But the remaining:
liver
does not provide the modern lexical base.
A mechanically applied prefix rule therefore creates:
a false etymology and false semantic decomposition.
This is different from a cranberry morpheme
The live eduKateSG cranberry-morpheme article discusses forms such as:
cran-
in:
cranberry.
There, a piece may survive only inside one or a few words.
A morphological illusion is different:
a familiar independent-looking piece appears to be present, but the segmentation is misleading.
This is also different from folk etymology
Folk etymology happens when speakers reinterpret an unfamiliar word through familiar material and the reinterpretation can affect:
Morphological illusion can occur:
during immediate word processing
without any historical language change.
So:
folk etymology → social/historical reanalysis.
morphological illusion → misleading structural cue in a word form.
This differs from rebracketing too
Rebracketing changes:
where speakers locate a boundary between words or morphemes.
A morphological illusion may never become conventional.
The reader simply entertains:
a false segmentation.
Why does this matter for vocabulary learning?
Morphology is one of the best tools for decoding unfamiliar words.
Students learn:
un- + happy → unhappy
and:
predict + -able → predictable.
That strategy is powerful.
But if students are taught:
“Always split any familiar-looking prefix or suffix,”
they will overanalyse.
The correct strategy is:
segment → test meaning → verify usage/history if needed.
Not every -er word contains productive -er
Transparent:
teach + -er → teacher.
Transparent:
farm + -er → farmer.
Misleading:
corner.
The sequence:
-er
can therefore be:
real suffix
or:
accidental ending.
Not every un- is the negative prefix un-
Students can make similar mistakes with initial sequences.
Compare a genuine negative prefix:
unhappy.
Now consider a word whose first letters merely resemble a familiar prefix.
The visual cue alone cannot prove:
morphological composition.
Meaning is a critical confirmation test
For:
teacher,
the meaning of the whole is compositionally linked to:
teach + person associated with action.
For:
corner,
the supposed:
corn + -er
analysis does not predict:
an angle or meeting point of surfaces/lines.
Semantic mismatch is a warning.
History can provide a second confirmation test
Some modern forms look decomposable because:
their spelling accidentally resembles current morphemes.
Etymology can show:
the word entered English through a different historical route.
But learners should use etymology carefully.
Historical origin explains:
where a word came from.
Current morphology explains:
how speakers represent and use it now.
Current morphological processing can differ from historical truth
A word can be historically simplex but:
look complex to modern readers.
The reading system may still react to:
apparent structure.
This is one reason morphological illusions are scientifically interesting:
processing can respond to form before complete lexical knowledge settles the analysis.
A 2026 study models this natural noise directly
Korochkina, Marelli and Rastle’s 2026 Journal of Memory and Language study asks how learners acquire derivational affixes from real text rather than from a perfectly cleaned training set.
The authors emphasise:
misleading forms such as deliver
as part of the natural input.
The model still learned important aspects of affix semantics from large-scale reading experience.
This means morphological learning tolerates noisy evidence
Real books contain:
Learners must infer:
which patterns are reliable.
They are effectively solving:
a statistical signal-versus-noise problem.
Frequency helps resolve illusions
If a learner repeatedly sees:
teach/teacher
farm/farmer
build/builder
with a consistent meaning relation, the agentive:
-er
pattern becomes strong.
But if:
corner
does not participate in the same semantic relation, the learner can store it as:
lexically whole.
Distribution matters more than one example
A single word can mislead.
A large set reveals:
the stable pattern.
This connects morphological learning to:
cross-situational and distributional learning.
Morphological illusions can slow dictionary inference
Suppose a student sees:
deliver.
If they wrongly infer:
de- = remove
and:
liver = organ,
the pieces do not help.
A student who notices the semantic mismatch should:
stop decomposing and switch strategy.
This is why morphology needs a confidence gate
A useful decoding sequence is:
False segmentation can create false definitions
Weak inference:
corner = someone/something that corns.
The morphological parts look familiar.
The semantic result is absurd.
That absurdity is useful evidence:
the segmentation is probably wrong.
But not all opaque morphology is fake morphology
Some genuinely complex words are semantically opaque.
Historical changes can weaken the transparency between:
parts
and:
whole.
Therefore:
semantic opacity does not automatically prove morphological illusion.
We need:
form + structure + lexical history + current evidence.
This connects to semantic transparency
The live eduKateSG article on semantic transparency asks:
How clearly can the meaning of a complex word be predicted from its parts?
This article asks:
Are those apparent parts even genuine morphemes in the first place?
So:
transparency → genuine structure, predictable or opaque meaning.
morphological illusion → apparent structure may be false.
This connects to affix productivity
The new productivity article asks:
How freely can a real affix create new words?
This article supplies a prerequisite:
make sure the apparent affix is real before measuring its productivity.
False positives contaminate the pattern.
Spelling encourages decomposition
English writing displays words as discrete strings.
Readers become sensitive to recurring letter sequences.
A suffix such as:
-er
appears often enough that:
any final er
can attract morphological attention.
Orthography therefore contributes to:
morpheme hypothesis generation.
Sound can create comparable illusions
Spoken language also contains:
recurrent sound sequences.
A listener can misidentify:
a familiar affix or boundary.
This connects morphological illusions to:
phonological parsing.
Multilingual learners can see different morphological cues
A Singapore learner may know:
Cross-language knowledge can help recognise:
roots and affixes.
It can also create:
false analogies.
Strong learners test:
the English structure itself.
Scientific vocabulary rewards careful decomposition
Technical English contains many:
That makes morphology highly valuable.
But it also raises the cost of:
inventing a structure that is not there.
AI systems make the same kind of mistake
A generative system can explain a word through:
plausible-looking chunks
even when the etymology or morphology is false.
This is especially dangerous because:
the explanation can sound elegant.
A useful public audit prompt is:
For every claimed prefix, suffix or root, verify that the segmentation is supported by modern morphological evidence or reliable etymology. Do not infer structure from spelling resemblance alone.
A practical morphological-illusion test
Take:
corner.
Proposed analysis:
corn + -er.
Ask:
If not:
apparent morphology is misleading.
A practical genuine-derivative test
Take:
farmer.
Ask the same questions.
Base:
farm.
Suffix:
-er.
Meaning:
person associated with farming.
Related forms:
farm, farming.
Strong genuine decomposition.
Parent-friendly practice
Write:
farmer
corner
teacher
brother.
Ask:
Which endings really behave like the same -er suffix?
Then ask:
Does the meaning of the whole follow from the proposed base?
The child learns:
spot pattern, then test it.
Teacher-friendly diagnosis
Letter-sequence fallacy: familiar letters treated as proof of a morpheme.
Prefix overanalysis: initial letters split from a base that does not exist semantically.
Opacity/illusion collapse: genuine opaque derivatives treated as pseudo-morphological.
Etymology-as-morphology error: historical parts assumed to be active modern morphemes.
Dictionary avoidance: absurd decomposition maintained rather than verified.
Clean-input myth: morphological learning assumed to happen without misleading forms.
Hilary Mantel as a high-level literary lens
A familiar-looking piece can seduce the reader into a story about the whole word.
The transferable lesson is:
recognition is not proof; a pattern should survive a meaning test before we trust it.
The value is disciplined observation, not imitation.
The larger lesson
Morphology is powerful because recurring pieces often carry recurring meaning.
But English contains:
false friends inside words.
Farmer really supports:
farm + -er.
Corner merely looks as though it does.
Deliver contains familiar-looking material that does not yield its modern meaning through naive decomposition.
The mature vocabulary learner asks:
Are these really morphemes, or do they only look like morphemes?
That is the morphological-illusion check.
Connections eduKateAI can learn
recurring letter/sound sequence → morpheme hypothesis
hypothesis → semantic + lexical + historical verification
apparent affix ≠ guaranteed real affix
pseudo-derived form → natural noise in morphological learning
genuine morphology → form-meaning relation survives verification
AI etymology → never derive origin from visual resemblance alone
The public principle is that morphological parsing is probabilistic: readers use recurring forms to hypothesise structure, but misleading lookalikes require semantic and lexical verification before the decomposition is accepted.
Related eduKateSG resources
- Morphological Productivity in English Vocabulary
- Cranberry Morphemes and Fossil Word Parts
- Semantic Transparency and Opacity
- Folk Etymology in English Vocabulary
- Rebracketing in English Vocabulary
Research basis
This article draws on Korochkina, Marelli and Rastle, “Morphemes in the wild: Modelling affix learning from the noisy landscape of natural text,” Journal of Memory and Language, Volume 148, April 2026, article 104746. The study explicitly models misleading forms such as deliver as part of the noisy natural input from which derivational-affix knowledge is learned. Source: https://doi.org/10.1016/j.jml.2026.104746.