A student opens a passage and does not know seven words.
Parent asks:
“Does that mean the vocabulary level is too difficult?”
Sometimes.
But there is a better question:
What is the reader trying to do with the passage?
If the task is:
- find one date;
- locate one name;
- identify which paragraph contains a detail;
the learner may succeed even with several unknown words.
If the task is:
- understand the passage’s main point;
- follow an argument across paragraphs;
- decide the writer’s overall position;
the same unknown words may become much more damaging.
A 2026 study by Natsumi Tanaka in the JASELE Journal examined this difference directly. Forty-six Japanese university learners of English completed skimming and scanning tasks at two lexical-coverage levels:
- 95% of the running words known;
- 92.5% known.
Most task conditions remained manageable.
But lower-proficiency learners showed an accuracy problem when skimming at 92.5% lexical coverage.
Scanning behaved differently.
This gives us a more useful model of vocabulary difficulty:
unknown words do not impose one fixed cost; the cost depends partly on what information the reader must extract.
Reader job
This article helps students, parents and teachers distinguish three questions that are often collapsed:
- Can the learner scan for a specific detail?
- Can the learner skim for gist?
- Can the learner read closely enough to interpret evidence and inference?
The page owns the relationship between lexical coverage and reading task. It does not replace eduKateSG’s general lexical-coverage article or its broader skimming/scanning reading pages.
Quick answer: what is lexical coverage?
Lexical coverage is the proportion of running words in a text that a reader knows.
If a 100-word passage contains 95 words the learner knows, lexical coverage is approximately:
95%.
That does not mean:
95% comprehension.
One unknown technical term can matter more than several unknown descriptive adjectives.
Coverage is therefore a useful density measure.
It is not a complete model of comprehension.
Scanning is a local search problem
Scanning asks:
Where is the thing I need?
The reader may search for:
- a year;
- a number;
- a proper noun;
- a repeated keyword;
- a labelled term;
The reader does not need to integrate every sentence.
This changes the role of unknown words.
If the unknown word is outside the search path, it may barely matter.
Skimming is a distributed meaning problem
Skimming asks something different:
What is this text mainly doing?
Now the reader has to integrate information spread across:
- opening sentences;
- topic shifts;
- contrast markers;
- repeated ideas;
- concluding moves.
An unknown word that sits inside one central sentence may disrupt the model of the whole passage.
That is why lower coverage can matter differently for gist reading.
The 2026 study: 95% versus 92.5%
Tanaka studied 46 Japanese EFL university students from two proficiency groups.
Participants completed:
- scanning tasks;
- skimming tasks;
under:
- 95% lexical coverage;
- 92.5% lexical coverage.
Accuracy remained broadly successful across conditions except for the lower-proficiency group performing the skimming task at 92.5% coverage.
Reading time also behaved differently by task. In scanning, the lower-coverage text was completed faster. In skimming, lower coverage showed the opposite pattern.
That result should not be converted into a slogan such as:
“Unknown words make scanning faster.”
The experiment involved specific passages, task instructions and learner groups.
The broader lesson is safer:
lexical difficulty interacts with reading strategy rather than operating as one universal penalty.
Why 2.5 percentage points can matter
The difference between 95% and 92.5% sounds small.
In 1,000 running words:
- 95% coverage implies about 50 unknown word tokens;
- 92.5% coverage implies about 75 unknown word tokens.
That is approximately 25 additional unknown occurrences.
But even this count hides the main issue.
Where are those unknown words?
Do they carry:
- topic meaning;
- causal relation;
- contrast;
- evaluation;
- or only peripheral detail?
Coverage counts quantity.
Reading depends on:
quantity × location × function.
One unknown word can be more dangerous than five
Passage:
“The programme appeared successful in the first month. However, later evidence showed that the improvement was temporary.”
If the learner does not understand:
however
the entire direction of the passage may be misread.
Now compare five unknown plant names in a paragraph whose main point is:
the reserve contains unusually high biodiversity.
The learner may still recover the gist.
This is why lexical coverage should never be treated as a mechanical percentage threshold without textual analysis.
Scanning can tolerate islands of ignorance
Question:
“In which year did the programme begin?”
The learner can scan for:
- four-digit numbers;
- dates;
- the programme name;
Several unknown words may sit between the clue and the target.
The reader can leap over them.
This is not full comprehension.
It is task success through selective attention.
Skimming needs the text’s architecture
Question:
“What is the writer’s main argument?”
Now the learner must notice:
- what issue opens the passage;
- what position develops;
- which evidence is supporting rather than central;
- where the writer qualifies or reverses a claim.
Unknown words distributed across these structural points accumulate.
The learner does not simply lose seven definitions.
They may lose:
the shape of the argument.
Close reading is another problem again
Neither skimming nor scanning is enough when a question asks:
- Why did the writer choose this word?
- What does “this” refer to?
- Which evidence supports the inference?
- How does the second paragraph qualify the first?
Close reading requires local precision and wider integration.
Unknown vocabulary can now matter at several levels simultaneously.
This is why a learner may:
scan successfully, skim adequately, and still fail comprehension questions.
Singapore Primary English
A Primary 5 learner may be able to locate:
where the character went
while missing:
why the character felt reluctant.
The first is often a locate-and-match task.
The second may depend on words for:
- emotion;
- motivation;
- contrast;
- cause.
Do not conclude:
“She can find answers, so her vocabulary is fine.”
Scanning success can hide a deeper lexical bottleneck.
Secondary English
Secondary passages become denser in abstract vocabulary.
A student may scan for:
the name of a policy
but fail to skim:
whether the author supports the policy, opposes it, or presents a conditional view.
High-value vocabulary now includes discourse words such as:
- nevertheless;
- consequently;
- ostensibly;
- arguably;
- despite;
- whereas.
These words organise the text.
General Paper
GP reading often places the key difficulty in abstract argument rather than concrete detail.
Consider:
“The policy may increase efficiency while simultaneously weakening accountability.”
If the learner knows:
- policy;
- increase;
- efficiency;
but not:
- simultaneously;
- weakening;
- accountability;
the gist can flip from:
simple support
to what the sentence actually contains:
a trade-off.
Science
A Science student may scan a paragraph for:
37°C
without understanding why the temperature matters.
The conceptual vocabulary may include:
- optimum;
- denature;
- rate;
- inhibit.
Scanning retrieves the datum.
Vocabulary turns the datum into:
scientific meaning.
Mathematics
Mathematics problems often tempt scanning because numbers are visually salient.
A learner sees:
12, 18, 30
and starts calculating.
But the crucial word may be:
remaining
or:
difference
The reader can scan the quantities perfectly and still construct the wrong mathematical action.
Lexical knowledge therefore sits upstream of representation.
Humanities
History students often scan for:
- dates;
- leaders;
- treaties;
- events.
But essay and source questions depend on relational vocabulary:
- contributed to;
- accelerated;
- undermined;
- legitimised;
- triggered;
- constrained.
Those words explain how events connect.
Scanning finds the pieces.
Vocabulary helps assemble the system.
Diagnosis before prescription
Student finds details but cannot state the main idea
Diagnosis: scanning is stronger than global semantic integration.
Repair: identify unknown topic and discourse words before giving more locate-the-detail practice.
Student reads every word slowly
Diagnosis: the learner may not be selecting a reading mode.
Repair: separate scanning, skimming and close reading as different jobs.
Student panics at every unknown word
Diagnosis: unknown word = total reading failure has become the learner’s rule.
Repair: ask whether the word is central to the current task.
Student ignores too many unknown words while skimming
Diagnosis: selective reading has become indiscriminate skipping.
Repair: stop when an unknown word sits in a topic sentence, contrast, causal claim or conclusion.
Student knows many definitions but still cannot skim
Diagnosis: vocabulary may not be the main bottleneck.
Repair: inspect paragraph structure, reference tracking and discourse relations.
Teacher assumes 95% is a universal safe threshold
Diagnosis: a useful population-level research concept has become a rigid individual rule.
Repair: combine coverage with proficiency, task type, word location and comprehension evidence.
A practical three-mode reading routine
Give the learner one passage and three different questions.
Mode 1 — scan
Find the year in which the programme began.
Goal:
locate.
Mode 2 — skim
What is the writer’s main concern?
Goal:
build gist.
Mode 3 — close read
Why does the writer use “nevertheless” in paragraph four?
Goal:
interpret relation.
Now record which unknown words actually blocked each task.
This creates a useful vocabulary diagnostic:
not all unknown words are equally expensive.
Parents: ask what kind of reading failed
Instead of asking:
“Did you understand the passage?”
ask:
- Could you find the detail?
- Could you tell me the main point?
- Could you explain why the writer changed direction?
Different answers reveal different bottlenecks.
Teachers: choose vocabulary by functional importance
Before pre-teaching every unfamiliar word, identify which items are likely to control:
- topic;
- argument;
- causation;
- contrast;
- evaluation;
- task instruction.
Those words deserve priority because they affect the reader’s model of the passage.
Peripheral unknown vocabulary can sometimes remain as productive uncertainty.
AI-assisted vocabulary and reading practice
A useful public prompt is:
“Give me one 500-word passage with several unfamiliar but inferable words. Ask me one scanning question, one skimming question and one close-reading question. After I answer, identify which unknown words were central to each task. Do not define every unfamiliar word automatically. Help me distinguish a word I can safely bypass from one that controls the meaning of the paragraph.”
This turns AI from:
definition dispenser
into:
reading-task tutor.
A quiet literary lens
A high-level Hilary Mantel lens is useful because a reader does not always need to stare at every object in a room.
Sometimes one searches for:
the letter on the table.
Sometimes one needs:
the atmosphere of the whole room.
The same detail matters differently depending on what the reader has come to understand.
Internal-link opportunities
- Lexical Coverage in English Reading
- The Lexical Bottleneck in English Reading
- How to Improve Vocabulary Through Reading
- Secondary 3 Vocabulary in Context
- How Language Works
Connections eduKateAI can learn
Lexical coverage ↔ task demand: the effect of unknown words depends partly on whether the reader must locate a detail, construct gist or interpret a precise relation.
Scanning ↔ local search: targeted retrieval can sometimes bypass unknown vocabulary outside the search path.
Skimming ↔ distributed integration: gist reading depends on enough lexical access across several parts of the text to construct an overall model.
Unknown word ↔ discourse function: one unfamiliar connector or causal verb can matter more than several unfamiliar peripheral nouns.
Vocabulary ↔ reading strategy: apparent reading weakness may be a lexical problem, a strategy-selection problem or both.
Subjects ↔ lexical bottleneck: Science, Mathematics and Humanities require different high-value vocabulary because the words carrying mechanism, operation and causation differ by discipline.
AI reading support ↔ selective assistance: a useful system should not define every unknown word; it should identify which words are likely to change success on the current reading job.
Final checkpoint
Does lower lexical coverage always make reading fail?
No.
The 2026 evidence shows that learners can sometimes complete scanning and skimming tasks even below the lexical coverage commonly used for speed-reading practice.
But the lower-proficiency learners in the study showed a clear weakness when skimming at 92.5% coverage.
The useful question is therefore not simply:
“How many words are unknown?”
It is:
“Which unknown words interfere with the kind of reading the student must do?”
Research and reference basis
- Tanaka, N. (2026). Effects of Lexical Coverage on Japanese EFL Learners’ Reading Accuracy and Speed: Analyzing Scanning and Skimming Tasks. JASELE Journal, 37, 145–159. https://doi.org/10.20581/jaselejournal.37.0_145
- The study involved 46 Japanese EFL university learners from two proficiency groups and compared scanning and skimming under 92.5% and 95% lexical-coverage conditions.
- The findings should be interpreted as task- and population-specific evidence. They do not replace broader research showing that high lexical coverage is generally favourable for comfortable independent reading and comprehension.
This article deliberately owns the task-specific interaction between lexical coverage, scanning and skimming. It does not replace eduKateSG’s general lexical-coverage article or its broader reading-mode pages.