How many words are in the sentence “The players played playful games”? The answer depends on what is being counted. There are five running words, four different written forms if repeated items are collapsed, and fewer lemmas or word families under particular rules. None of those counts tells us, by itself, how many meanings a learner can understand or use.
This guide separates tokens, types, lemmas and word families, then connects each unit to a learner question. It is not a universal vocabulary test or a fixed rule for every language. Morphology, spelling, meaning, register and learner knowledge complicate the boundaries. The worked examples are original and deliberately small so every grouping can be checked.
This specialist guide belongs under The Gold Standard of Vocabulary. It connects to The Gold Standard of Education: Learning That Transfers into Life and the verified How X Works hub.
eduKateSG · Vocabulary
Find your next learning step
Choose the lexical unit from the learner question, preserve the mapping from tokens to families and test the forms a learner actually understands.
ROUTE 1 · CHAPTERS 1–10
Define the counting units
Separate tokens, types, lemmas, families, inflection, derivation and coverage.
ROUTE 2 · CHAPTERS 11–20
Handle lexical complexity
Keep phrases, languages, tools, learner evidence, cues and frequency visible.
ROUTE 3 · CHAPTERS 21–31
Audit the learning claim
Test coverage, denominators, classroom use and common misconceptions.
ROUTE 4 · CHAPTERS 32–42
Extend the lexical account
Handle names, variants, compounds, affixes, tools, spelling and sampling.
ROUTE 5 · CHAPTERS 43–52
Transfer and continue
Compare depth, growth, subject use, unseen text and verified routes.
Chapter contents
Define the counting units · Chapters 1–10
1. Begin with the reader’s actual question
3. Count types without calling them meanings
5. Define a word family cautiously
6. Work through one complete count
7. Keep inflection and derivation separate
8. Do not assume family knowledge from one correct answer
Handle lexical complexity · Chapters 11–20
11. Keep homographs and part of speech visible
12. Handle multiword expressions
13. Check the language and morphological system
14. Compare tools before trusting the total
15. Build a learner-specific family check
16. Separate receptive and productive knowledge
18. Use frequency without confusing it with importance
Audit the learning claim · Chapters 21–31
21. Work a changed-context example
22. Make the denominator visible in vocabulary size
23. Design a practical classroom audit
24. Correct common misconceptions
25. Questions readers often ask
26. Recheck with an unseen text
27. Check whether a family relation helps comprehension
28. Preserve uncertainty in manual coding
29. Build a learner route from the count
Extend the lexical account · Chapters 32–42
32. Separate frequency bands from mastery bands
33. Examine false friends across languages
34. Measure retention across delay
35. Analyse a longer fictional passage
36. Use changed-context transfer as the final check
37. Mark proper nouns and named entities
39. Treat compounds explicitly
40. Separate morphological transparency from semantic transparency
Transfer and continue · Chapters 43–52
44. Handle learner spelling errors carefully
45. Use sampling to estimate a large inventory
46. Compare breadth with depth
47. Track growth without changing units
49. Apply the method to subject vocabulary
50. Create a transparent reporting card
CHAPTER 1 OF 52 · Define the counting units
1. Begin with the reader’s actual question
A corpus analyst may ask how often forms occur. A teacher may ask which meanings block a passage. A test designer may estimate receptive breadth. A writer may ask whether a learner can retrieve an accurate word for a new sentence. These questions need different evidence.
State the question before choosing the counting unit. “How many spaces-separated items are in the text?” can use a token rule. “How many base entries appear?” may use lemmas. “How many derivational families can this learner understand?” requires both a family rule and evidence about the learner. A convenient number should not quietly replace the question.
A token is one occurrence under the selected tokenisation rule. In “The players played playful games,” a simple space-based count gives five tokens: The, players, played, playful and games. If the sentence repeats games, the new occurrence adds another token.
Token counts describe running text length and frequency. They do not collapse repetition and do not directly measure vocabulary breadth. Punctuation, contractions, hyphens, numbers and multiword expressions require explicit rules. Software can count consistently while still using a rule unsuited to the language or task.
CHAPTER 3 OF 52 · Define the counting units
3. Count types without calling them meanings
A type is a distinct token form after stated normalisation. Lowercasing may merge The and the. Keeping case may separate them. In the example, all five forms are distinct, so there are five types under a case-insensitive spelling rule.
Types reduce repetition but still separate players from player and played from play. They can support text-complexity and lexical-diversity work when length and method are controlled. A type is not a meaning: bank as a financial institution and bank beside a river can remain one spelling type while expressing different senses.
A lemma commonly groups a base form with its inflectional variants that belong to the same part of speech, such as play, plays, played and playing as a verb lemma. Exact conventions vary by dictionary, tagger and language. Derived forms such as player or playful are generally separate lemmas.
Under a simple teaching rule, the sentence contains the verb lemma play through played, the noun lemma player through players, the adjective lemma playful, the noun lemma game and the function word the: five lemmas. Lemmas reduce predictable inflectional variation without assuming that every derived word is known.
A word family can include a base word, inflections and selected transparent derivations, such as play, played, player and playful. Family membership depends on morphological rules and the proficiency level assumed. A broad family can be useful for profiling texts while overestimating what an individual learner understands.
If the example groups played, players and playful under PLAY, the sentence might contain three families: THE, PLAY and GAME. That count is mathematically consistent with the rule. It does not prove that a learner who knows play understands playful or player.
Use the fictional line: “A careful planner replanned the plans carefully.” Under a simple space-and-punctuation rule it has seven tokens. Case-normalised written types are A, careful, planner, replanned, the, plans and carefully—seven types.
A lemma-based analysis might group replanned with replan and plans with plan, while planner, careful and carefully remain separate lemmas under the chosen convention. A broad family analysis might group plan, planner and replan together, and careful with carefully. Every reduction reflects a decision. Present the mapping table rather than only the final total.
CHAPTER 7 OF 52 · Define the counting units
7. Keep inflection and derivation separate
Inflection usually changes grammatical form without creating a new core lexeme: walk, walks, walked and walking. Derivation can create a new word or meaning relationship: walkable, walker or walkway. Languages differ, and even English boundaries can be debated.
For learners, the distinction matters because knowledge of a familiar inflection may transfer more reliably than knowledge of a derived form. Recognising avoid does not guarantee understanding avoidance or unavoidable. Test the derived item when that knowledge matters instead of awarding it automatically through the base.
CHAPTER 8 OF 52 · Define the counting units
8. Do not assume family knowledge from one correct answer
A receptive test item may sample the base word. A correct response supports knowledge under that item’s cue and options. It does not directly test every family member, every sense, productive use or register.
Laufer and Cobb describe the debate: family-based size tests may assume knowledge of derived forms from knowledge of a base, while critics argue that lemmas provide a safer unit. The practical lesson is not that families are always wrong. It is that the inference must match the test and learner evidence.
CHAPTER 9 OF 52 · Define the counting units
9. Distinguish text coverage from learner knowledge
Text coverage asks what proportion of running tokens belong to a selected list or known set. A family-based list can classify many forms through one entry. Learner knowledge asks whether the actual reader can understand those forms in context.
If ten derived tokens are assigned to families whose base words a learner knows, a program may report them as covered. If the learner cannot interpret the derivations, effective coverage is lower. Label the list unit and avoid presenting computational coverage as a direct observation of comprehension.
One lemma can carry several related senses. A learner may know bright as “giving much light” but not bright as “intelligent” or “promising.” Lemma counts do not split those senses unless the method explicitly does.
Test the meaning used in the task. A dictionary entry offers possible senses; context selects among them. Counting one known lemma should not license every sense. Conversely, partial sense knowledge is not zero knowledge. Record which meaning the learner demonstrated.
CHAPTER 11 OF 52 · Handle lexical complexity
11. Keep homographs and part of speech visible
Written forms can hide unrelated meanings or different grammatical roles. The noun record and verb record share spelling but differ in stress and use. Some lemmatisers separate them through part-of-speech tags; others may not.
Report the tagging convention. If a learner recognises record as a noun, do not infer confident use as a verb. A type count cannot see this distinction, and a family count may broaden it further. The unit is a tool, not a complete model of lexical knowledge.
Phrases such as take part, in spite of and make a decision carry conventional relationships that single-word counts can miss. A learner may know each component yet misunderstand the expression, or know the phrase without being able to explain every separate word.
Decide whether the task treats the expression as one unit, several tokens or both. Preserve both views when useful. Token statistics can remain word-based while teaching records flag the expression. Do not force every language pattern into a one-word family model.
CHAPTER 13 OF 52 · Handle lexical complexity
13. Check the language and morphological system
English-based assumptions do not transfer unchanged to languages with richer inflection, compounding, clitics or non-spaced writing. A tokenizer and lemmatiser need language-appropriate rules. Even within one language, genres and spelling conventions can change segmentation.
Document the tool, model and version used. Manually inspect a sample, especially the forms driving a decision. An automated lemma count is reproducible only if another person can recover the same processing choices.
CHAPTER 14 OF 52 · Handle lexical complexity
14. Compare tools before trusting the total
Run a small passage through two analysers and inspect disagreements. One may treat cannot as one token while another splits can and not. One may tag planning as a noun, another as a verb. Different totals can both follow their internal rules.
Resolve differences by returning to the reader question. Do not average the two counts. Choose or adapt a rule, record it and keep ambiguous items visible. Agreement on a headline total is less important than transparent treatment of consequential forms.
CHAPTER 15 OF 52 · Handle lexical complexity
15. Build a learner-specific family check
Suppose Mei understands create and created. Ask her to explain creative, creation and recreate in short contexts. She correctly explains creation, confuses creative with “making something now” and has not seen recreate. Her demonstrated family knowledge is uneven.
A useful record shows each form, context, response and next teaching step. Do not award the whole CREATE family or mark the whole family unknown. Teach the missing relationships, then retest with changed contexts. The family becomes a learning map rather than a scoring shortcut.
CHAPTER 16 OF 52 · Handle lexical complexity
16. Separate receptive and productive knowledge
A learner may select the meaning of reluctant from options but not retrieve it when speaking or writing. They may recognise a derived form after seeing the base while being unable to construct it accurately.
State the mode tested: recognition, recall, explanation or production. Productive knowledge needs an opportunity to express a suitable word without the answer being supplied. Receptive success remains valuable; it simply supports a different claim.
A translation, definition, sentence blank and picture provide different cues. Scores can change because the cue changes, not only because knowledge changes. The existing specialist article Comparing Test Scores When the Cues Change develops that issue directly.
When comparing counts based on tests, keep cue type, response format, time and scoring rule aligned. A family total built from prompted base words cannot be compared casually with independent production of derived forms.
CHAPTER 18 OF 52 · Handle lexical complexity
18. Use frequency without confusing it with importance
Frequency counts depend on corpus, unit and normalisation. Lemmatization combines inflected forms, which can change rank compared with raw written types. A frequent family may contain a rare derivative that a learner has never encountered.
Use the frequency list matching the population and task. Keep proper nouns, formulae and specialist terms under explicit rules. A low-frequency word may still be essential for a science lesson or a learner’s immediate purpose. Frequency guides selection; it does not replace curricular judgment.
CHAPTER 19 OF 52 · Handle lexical complexity
19. Connect dispersion with the chosen unit
A word can occur many times in one text and nowhere else. Dispersion shows how broadly it appears across documents or sections. Whether forms are combined into a lemma or family changes the dispersion calculation.
The specialist guide Calculating Word Dispersion Across Unequal Text Lengths compares raw counts, document coverage and length-aware checks. Use the same lexical unit across documents or explain the change; otherwise apparent spread may come from grouping rules.
CHAPTER 20 OF 52 · Handle lexical complexity
20. Do not turn coverage into guaranteed comprehension
High lexical coverage makes reading easier on average, but comprehension also depends on syntax, background knowledge, cohesion, inference and purpose. Unknown words can cluster in a decisive sentence. Familiar words can express an unfamiliar concept.
The specialist guide Bounding Unfamiliar-Word Runs When Overall Coverage Is Fixed shows how equal overall coverage can hide different local runs. Report coverage as one condition, not a promise of understanding.
Consider “The observers observed an unobservable change.” There are six tokens under a space rule. Observers, observed and unobservable share some form of relation to observe, but they play different grammatical and semantic roles. A broad OBSERVE family count may group all three.
Now ask a learner to explain each. If they know observed, interpret observers as “things being watched” and cannot explain unobservable, the family-level classification overstates demonstrated knowledge. The changed case tests whether the reader can apply the method beyond PLAY and CREATE.
CHAPTER 22 OF 52 · Audit the learning claim
22. Make the denominator visible in vocabulary size
A test score of 80% means little without the sampled items, frequency bands, lexical unit and extrapolation rule. Twenty correct base-word items do not literally equal twenty complete families. Scaling a sample to thousands of items adds sampling and construct assumptions.
Report raw correct items, tested unit, score, uncertainty where appropriate and the population of words the test claims to represent. Avoid converting a score into a precise inventory of everything the learner knows.
Choose a short relevant passage. Count tokens with a declared rule. Identify types that impede the task. Map selected forms to lemmas and candidate families. Ask the learner about the actual forms and meanings in context. Add a changed-form and changed-context check.
Finish with a small teaching decision: clarify a sense, practise an inflection, teach a derivation, record a phrase or revisit retrieval. The purpose is better learning, not the largest possible vocabulary total. Keep the record light enough to use again.
A token is not a type. A type is not a lemma. A lemma is not automatically a full word family. A known base does not guarantee known derivatives. One spelling does not guarantee one meaning. A list match does not guarantee comprehension or production.
The reverse cautions matter too. Family grouping is not inherently invalid; it can be useful for text profiling and research when the assumptions fit. Lemmas do not solve every sense and phrase problem. Choose the unit for the claim and keep its limits visible.
After the learner works through a family in one passage, select an unseen text containing some familiar forms and one new related form. Ask for meaning, grammatical role and evidence from context before offering help. The task shows whether morphology and meaning transfer beyond the original examples.
Record partial success. A learner may infer the general meaning while missing register or part of speech. Repair that specific gap and return to another unseen example later. The aim is not to trap the learner but to establish what can now be done independently.
This final check also tests the counting assumption. If broad family grouping predicted knowledge that repeatedly fails in unseen contexts, use a narrower unit for that learner and purpose. If the learner handles transparent derivatives reliably, the family may be an efficient planning tool. Evidence, rather than habit, determines the grouping.
CHAPTER 27 OF 52 · Audit the learning claim
27. Check whether a family relation helps comprehension
Morphological relatedness is useful only when the learner can activate it. Present a base and derived form in context, ask for the relationship and request a reason. A correct guess without explanation may be fragile.
Compare transparent pairs such as teach–teacher with less transparent ones such as history–historical. Record which affixes and spelling changes the learner handles. The result informs teaching more directly than a universal family size.
CHAPTER 28 OF 52 · Audit the learning claim
28. Preserve uncertainty in manual coding
Human analysts can disagree about compounds, senses and family membership. Use a coding guide, double-code a sample and record adjudication. Agreement does not prove the rule is ideal, but it shows whether it can be applied consistently.
Keep an “uncertain” category rather than forcing every item. Report how uncertain forms affect totals. A range can be more honest than one apparently exact count.
CHAPTER 29 OF 52 · Audit the learning claim
29. Build a learner route from the count
For repair, teach the base meaning and one high-value form. For stabilisation, compare forms across sentence roles. For extension, examine less transparent derivations, register and subject uses.
The route should end with independent reading or writing, not a bigger spreadsheet. Counting diagnoses the next step; accurate, flexible use is the educational outcome.
CHAPTER 30 OF 52 · Audit the learning claim
30. Distinguish headwords from lexical knowledge
A dictionary headword organises entries. Its structure reflects editorial conventions and may group or split senses differently from a corpus lemma or teaching family. Counting headwords therefore does not directly count everything a learner knows.
When a project uses a dictionary, name the edition and entry rule. Check phrasal verbs, compounds and homographs. A learner may know a common sense without knowing the technical senses listed under the same headword. Use headwords for navigation and reference, then test the meaning required by the task.
CHAPTER 31 OF 52 · Audit the learning claim
31. Deal with conversion between parts of speech
English can use the same form as different parts of speech without an overt affix: to email and an email, to bottle and a bottle. Some analyses treat these as related lemmas; others split them by part of speech.
For teaching, ask whether the learner controls each role and sentence pattern. Recognising the noun email does not guarantee accurate verb tense or argument structure. State whether the count is form-based or part-of-speech sensitive. This prevents a stable spelling from hiding a new grammatical demand.
CHAPTER 32 OF 52 · Extend the lexical account
32. Separate frequency bands from mastery bands
A word family can belong to a high-frequency list while the learner’s knowledge is partial. Frequency is a property of the reference corpus and counting method; mastery is a property inferred from learner performance.
Keep two fields. One records corpus band, the other records observed knowledge under tasks. Do not promote a learner to “5,000 words known” merely because sampled items come from that band. Describe the estimate and observed gaps. Teaching can then prioritise frequent missing forms without turning the list into a label.
CHAPTER 33 OF 52 · Extend the lexical account
33. Examine false friends across languages
Forms that resemble a known word in another language can invite an incorrect meaning. A cognate may help, while a false friend can mislead. Family counts within English do not capture this cross-language relationship.
Invite learners to state the inferred link and verify it. Record the language and meaning rather than discouraging multilingual reasoning. Cross-language knowledge is a resource when checked. A transparent report distinguishes evidence from resemblance.
A family check immediately after instruction shows supported learning under fresh memory. Repeating selected forms days or weeks later tests retention under changed timing. Include both familiar and new contexts.
Keep the lexical unit constant between checks. If the first test presents bases and the second requires derived production, a lower score may reflect both delay and task change. A useful design varies one dimension at a time, then deliberately combines dimensions for transfer after foundations are stable.
CHAPTER 35 OF 52 · Extend the lexical account
35. Analyse a longer fictional passage
Suppose a 120-token passage contains 86 written types, 74 lemmas under a part-of-speech-sensitive analyser and 68 broad families. A learner is directly tested on 20 forms and explains 15 accurately. The four corpus totals describe the passage; the 15 responses describe sampled learner performance.
It would be wrong to say the learner knows 68 families because the passage has 68 families, or that 15 correct items prove 75% of every form in the passage. Use the data to choose follow-up checks. If missed items cluster in derivations, teach morphology; if they cluster in senses, teach contextual distinctions.
CHAPTER 36 OF 52 · Extend the lexical account
36. Use changed-context transfer as the final check
After teaching observe, observer, observation and observant, give a new task about a sports match rather than the original science experiment. Ask the learner to select and explain the form. Transfer requires preserving the word relationship while changing the setting.
Score meaning, grammar and fit separately. A grammatically correct sentence can still misuse the concept. The final report should say what transferred and what needs repair. This mirrors the broader Gold Standard principle: useful knowledge survives a changed context without losing its limits.
CHAPTER 37 OF 52 · Extend the lexical account
37. Mark proper nouns and named entities
Names such as Singapore, Monday and Mei can inflate type counts while serving a different role from general vocabulary. Some profiling systems exclude proper nouns; others count them or treat them as inferable. The choice affects coverage.
Define named-entity rules before analysis. A learner may recognise a name through world knowledge without knowing its cultural or grammatical use. Conversely, an unfamiliar technical name can block meaning. Preserve the token in the text record even if it is excluded from a vocabulary-size estimate.
Colour and color may be separate written types but variants of one lexical item for many learning purposes. Programme and program overlap in some contexts but not all. Historical and regional spellings require audience-aware treatment.
Normalising variants can improve comparability while hiding register or regional expectations. Keep original forms and a normalised field. Teach the version needed for the learner’s context without describing other standard forms as universally wrong. A type count should state whether variants were merged.
School bus contains two orthographic words and one conventional concept. Blackboard can be one written word, while black board may describe any board that is black. Hyphens add another boundary. Tokenisation alone cannot decide lexical status.
Use corpus, dictionary and task evidence. A learner who knows school and bus may still need the compound’s conventional meaning, but transparent compounds differ from idioms. Record compounds when they matter rather than forcing every combination into a family.
CHAPTER 40 OF 52 · Extend the lexical account
40. Separate morphological transparency from semantic transparency
Unhappiness has visible parts that often support inference from happy. Department shares a historical relation with depart but its modern meaning is not derived transparently for a learner. Family grouping based only on spelling can create false relationships.
Ask whether the learner can use the affix rule and whether the resulting meaning fits the context. Teach opaque items as separate lexical knowledge where needed. Morphological analysis is evidence, not an automatic decoder.
Some affixes combine freely with many bases; others are limited, technical or historically fixed. A learner who understands -ness in kindness may infer darkness but not every apparent ending. Spelling changes can further complicate the relationship.
Build affix knowledge through several clear examples and non-examples. Avoid telling learners that removing a familiar ending always reveals a base. Test the proposed base in a reliable dictionary and context. Productive knowledge includes knowing when a pattern should not be extended.
CHAPTER 42 OF 52 · Extend the lexical account
42. Distinguish a pedagogical family from a corpus family
A corpus list may group forms for frequency analysis at a defined affix level. A teacher may use a smaller temporary family containing only forms needed this week. Both can be legitimate if labelled.
Do not import a broad research family directly into a beginner’s score. Instead, map the list family to demonstrated items. The pedagogical family can expand as the learner acquires morphology and meanings. This makes grouping responsive to proficiency rather than fixed by software convenience.
Automated lemmatisers infer base forms from spelling and context. They can mis-tag ambiguous words, unfamiliar names and learner errors. A wrong part-of-speech tag can send a token to the wrong lemma.
Inspect a sample and every high-impact item. Record confidence or correction rules. Do not silently hand-correct only the examples that support the desired result. Reproducibility requires a correction log and preserved original output.
CHAPTER 44 OF 52 · Transfer and continue
44. Handle learner spelling errors carefully
A misspelling creates a new written type in raw text, but counting it as new vocabulary would misinterpret the learner’s production. At the same time, automatically correcting it can hide an important spelling problem.
Keep original form, intended form where reasonably established, and correction status. For lexical diversity, report the normalised method. For teaching, address meaning, word choice and spelling separately. When intention is unclear, ask rather than guess.
CHAPTER 45 OF 52 · Transfer and continue
45. Use sampling to estimate a large inventory
Testing every possible word is impractical. Vocabulary tests sample items from frequency bands or lists and extrapolate. The estimate depends on representative sampling, item quality and scoring.
Report confidence intervals or score ranges when available. One unusual item can move a small test substantially. Do not convert an estimated 5,000-family score into a statement that exactly 5,000 complete families are known. The figure is a model-based summary.
Breadth asks how many lexical items are known under a criterion. Depth asks about senses, collocations, grammar, register, morphology and associations. A family count mostly supports a breadth question and cannot stand in for every depth dimension.
Pair a small breadth sample with tasks that require explanation and use. A learner can have broad recognition and shallow control, or deep command of a smaller specialist vocabulary. Choose instruction from the observed profile rather than one ranking.
A baseline counted lemmas while a later test reports word families. The second number may be smaller despite genuine learning because the unit changed. A new corpus or test form can create another discontinuity.
Use parallel forms, common anchor items and the same scoring construct where possible. When the instrument changes, establish a bridge study or begin a new series. Do not splice incomparable scores into one smooth growth chart.
When a learner misses a derived word, identify whether the problem is base meaning, affix meaning, spelling change, part of speech, sense selection or retrieval. Different errors require different repairs.
For disagreement, a learner might know agree and the negative prefix dis- but not understand the noun suffix -ment. Teach the specific relationship, then test disagreement in a new sentence and another transparent -ment noun. This is more useful than repeating the family name.
CHAPTER 49 OF 52 · Transfer and continue
49. Apply the method to subject vocabulary
In science, react, reaction, reactive and reactivity share a family relationship but express different concepts and grammatical roles. Knowing react does not guarantee correct interpretation of reactivity. In mathematics, equal, equality and equate also differ.
Use subject definitions and authentic sentence patterns. Ask the learner to choose the form that makes the claim accurate. Subject vocabulary reveals why morphology and concept knowledge must work together.
A useful report might show 120 passage tokens, 86 types, 74 lemmas under Tool X version Y, 68 broad families under Rule Z, and 15 target forms directly tested. It would then state which 11 were understood in context and which four need teaching.
This card prevents the largest number from becoming the only story. It separates text structure, automated grouping and learner evidence. Another teacher can continue from it without assuming that 68 families were fully mastered.
Which count is the correct number of words?
There is no context-free answer. Tokens, types, lemmas and families answer different questions under defined rules.
Should young learners use word families?
They can learn transparent relationships, but teachers should check each form and avoid assuming that the whole family is known.
Does knowing played mean knowing play?
Often the relationship is accessible, but demonstrate it in the learner and task rather than treating every inflectional inference as certain.
Can software measure vocabulary knowledge?
It can analyse forms and responses under a model. Learner knowledge still requires valid tasks, scoring and interpretation.
CHAPTER 51 OF 52 · Transfer and continue
51. Continue through the verified vocabulary routes
Return to The Gold Standard of Vocabulary for meaning, context, retrieval and transfer. Read Comparing Test Scores When the Cues Change for assessment conditions, Calculating Word Dispersion Across Unequal Text Lengths for distribution, and Bounding Unfamiliar-Word Runs When Overall Coverage Is Fixed for local difficulty hidden by overall coverage.
Laufer and Cobb, How Much Knowledge of Derived Words Is Needed for Reading? reviews family-based counting, lemma alternatives, derived-word knowledge and text coverage. Its findings and debates support the distinction between a useful counting unit and demonstrated learner knowledge. The sentences, counts, learner responses and teaching sequence in this guide are original examples. Apply language-specific morphology and the documentation of any analyser used.