SEMANTIC PREFERENCE · COLLOCATION · SEMANTIC PROSODY · CORPUS LINGUISTICS · CO-SELECTION · WORD PATTERNS · MEANING CLASSES
Semantic preference is the tendency of a word or phrase to occur with members of a recurring semantic class. A verb such as pose may repeatedly select nouns referring to risks, threats and challenges. The individual collocates differ, but they share a meaning category.
This differs from collocation, which focuses on particular word partnerships, and from semantic prosody, which concerns the evaluative or attitudinal colouring associated with recurrent contexts. Semantic preference asks a more abstract question: what kinds of meanings tend to fill this slot?
This guide explains semantic preference, co-selection, collocation, colligation, semantic prosody, corpus concordances, verb-object preferences, register effects and semantic classes. Existing eduKateSG Collocation and Semantic Prosody articles remain untouched.
Collocation tells us which words meet. Semantic preference tells us what kinds of meanings keep arriving.
2. Semantic preference
Semantic preference is the tendency of a lexical item or phrase to occur with words from a recurring semantic class, such as problems, quantities, human roles, processes or evaluation.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
3. Beyond individual collocation
Collocation links particular lexical items; semantic preference abstracts across several collocates to identify the meaning class they share.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
4. Different from semantic prosody
Semantic prosody concerns evaluative or attitudinal colouring that emerges from recurrent context. Semantic preference concerns the semantic category of neighbours and does not have to be positive or negative.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
5. Corpus evidence
Semantic preference is discovered by examining many authentic concordance lines and grouping recurrent collocates by meaning.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
6. Co-selection
Words are often selected together with lexical, grammatical and semantic patterns. Semantic preference is one part of a larger phraseological unit of meaning.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
7. Register effects
A word can show different semantic preferences in academic, journalistic, technical or conversational registers.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
8. Sense effects
Different senses of a polysemous word can have different semantic preferences. Corpus analysis should therefore separate senses where possible.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
9. Grammatical position
Preference may occur in subject, object, complement or modifier positions. The semantic class is tied to a structural slot, not merely to nearby words in general.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
10. Verb-object preference
Many verbs show recurrent semantic classes in their objects: address often takes issues/problems/concerns; conduct often takes research/investigations/studies.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
11. Adjective-noun preference
Adjectives can prefer nouns from semantic classes, such as acute with problems/conditions/shortages in particular registers.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
12. Noun-modifier preference
Nouns can attract modifiers from recurring semantic classes, revealing how discourse repeatedly frames the concept.
The key is abstraction: several different lexical collocates may reveal one recurring semantic preference when they belong to the same conceptual class.
13. Prepositional preference
A lexical item can show semantic tendencies inside a following prepositional phrase, such as causes, beneficiaries or domains.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
14. Semantic sets
Useful semantic classes include people, institutions, quantities, problems, benefits, processes, emotions, evidence, change, communication and evaluation.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
15. Preference is probabilistic
A semantic preference is a tendency, not a grammatical ban. Words can occur outside their usual semantic neighbourhood.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
16. Preference and naturalness
Learners who know a word’s semantic preferences can choose combinations that sound more conventional and contextually plausible.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
17. Preference and comprehension
Semantic expectations help readers narrow likely interpretations and anticipate what kind of information may follow.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
18. Preference and writing
Writers can use semantic preference to check whether a chosen verb or adjective normally selects the kind of noun they have supplied.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
19. Preference and translation
A translation equivalent may share core denotation while preferring different semantic classes in the target language.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
20. Preference and language learning
Teaching a word with several collocates from the same semantic class can reveal a deeper usage pattern than memorising one phrase.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
21. Preference and dictionaries
Advanced learner dictionaries and corpus tools can reveal recurring patterns indirectly through examples and collocations even when the label semantic preference is not used.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
22. AI-era preference
AI can suggest plausible classes, but corpus verification remains important because fluent generation can overgeneralise patterns or invent typicality.
Semantic preference is probabilistic and corpus-dependent. A pattern should be verified across many examples and interpreted relative to register, sense and grammatical position.
23. Semantic-preference casebook — Cases 1–20
1. address
Illustrative semantic class: problem or topic entities. Possible members: issues/problems/concerns/questions. Examples: address a concern; address the issue; address the question.
Corpus question: collect concordance lines for address and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
2. pose
Illustrative semantic class: potential difficulty or discourse challenge. Possible members: risk/threat/challenge/problem/question. Examples: pose a risk; pose a threat; pose a challenge.
Corpus question: collect concordance lines for pose and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
3. raise
Illustrative semantic class: discourse or public-attention items. Possible members: question/issue/concern/awareness. Examples: raise a question; raise concerns; raise awareness.
Corpus question: collect concordance lines for raise and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
4. meet
Illustrative semantic class: standards, needs or goals. Possible members: need/demand/requirement/target/deadline. Examples: meet a requirement; meet demand; meet the deadline.
Corpus question: collect concordance lines for meet and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
5. reach
Illustrative semantic class: endpoints or achieved states. Possible members: agreement/conclusion/decision/target/destination. Examples: reach an agreement; reach a conclusion; reach the target.
Corpus question: collect concordance lines for reach and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
6. draw
Illustrative semantic class: cognitive or discourse operations. Possible members: conclusion/distinction/comparison/attention. Examples: draw a conclusion; draw a distinction; draw attention.
Corpus question: collect concordance lines for draw and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
7. conduct
Illustrative semantic class: formal organised activities. Possible members: research/study/survey/investigation/interview. Examples: conduct research; conduct a survey; conduct an investigation.
Corpus question: collect concordance lines for conduct and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
8. perform
Illustrative semantic class: deliberate activities or functions. Possible members: task/test/function/operation/procedure. Examples: perform a task; perform a test; perform a function.
Corpus question: collect concordance lines for perform and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
9. provide
Illustrative semantic class: resources or enabling inputs. Possible members: information/support/service/evidence/access. Examples: provide information; provide support; provide access.
Corpus question: collect concordance lines for provide and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
10. offer
Illustrative semantic class: benefits, resources or possibilities. Possible members: help/advice/service/opportunity/choice. Examples: offer advice; offer an opportunity; offer help.
Corpus question: collect concordance lines for offer and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
11. gain
Illustrative semantic class: acquired resources or positive states. Possible members: experience/access/insight/confidence/support. Examples: gain experience; gain access; gain insight.
Corpus question: collect concordance lines for gain and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
12. acquire
Illustrative semantic class: things obtained or learned. Possible members: knowledge/skill/language/property/company. Examples: acquire knowledge; acquire a skill; acquire a company.
Corpus question: collect concordance lines for acquire and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
13. develop
Illustrative semantic class: emerging capabilities, systems or conditions. Possible members: skill/system/method/strategy/disease. Examples: develop a skill; develop a system; develop a condition.
Corpus question: collect concordance lines for develop and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
14. establish
Illustrative semantic class: entities, relations or propositions made stable. Possible members: relationship/system/fact/principle/company. Examples: establish a relationship; establish a system; establish the facts.
Corpus question: collect concordance lines for establish and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
15. maintain
Illustrative semantic class: states or arrangements kept stable. Possible members: standard/system/relationship/pressure/balance. Examples: maintain standards; maintain a relationship; maintain balance.
Corpus question: collect concordance lines for maintain and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
16. achieve
Illustrative semantic class: desired endpoints or states. Possible members: goal/result/outcome/success/balance. Examples: achieve a goal; achieve results; achieve balance.
Corpus question: collect concordance lines for achieve and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
17. obtain
Illustrative semantic class: formally acquired resources or outcomes. Possible members: permission/information/result/access/evidence. Examples: obtain permission; obtain information; obtain evidence.
Corpus question: collect concordance lines for obtain and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
18. submit
Illustrative semantic class: formal documents or claims. Possible members: application/report/request/proposal/evidence. Examples: submit an application; submit a report; submit evidence.
Corpus question: collect concordance lines for submit and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
19. issue
Illustrative semantic class: official communicative documents/actions. Possible members: warning/statement/report/guidance/order. Examples: issue a warning; issue guidance; issue a statement.
Corpus question: collect concordance lines for issue and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
20. launch
Illustrative semantic class: initiated public or organised activities. Possible members: campaign/product/service/inquiry/programme. Examples: launch a campaign; launch a product; launch an inquiry.
Corpus question: collect concordance lines for launch and inspect the relevant grammatical slot. Do multiple collocates cluster into the proposed class, or is the impression based on a few memorable examples?
Collocation distinction: identify one strong individual collocate and then step upward to the broader semantic class. This separates lexical partnership from semantic preference.
Prosody distinction: decide whether the class also carries evaluative colouring. If it does, keep the evaluative tendency separate from the descriptive semantic category.
Learning use: store several examples from the same class so the learner can generalise the usage pattern beyond one phrase.
24. Semantic-preference casebook — Cases 21–40
21. deliver
Illustrative semantic class: outputs, performances or transferred items. Possible members: service/result/speech/message/package. Examples: deliver a service; deliver results; deliver a speech.
Sense check: determine which sense of deliver is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with deliver. Then verify with corpus evidence rather than intuition alone.
22. generate
Illustrative semantic class: produced outputs. Possible members: data/revenue/interest/electricity/idea. Examples: generate data; generate revenue; generate interest.
Sense check: determine which sense of generate is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with generate. Then verify with corpus evidence rather than intuition alone.
23. produce
Illustrative semantic class: created outputs. Possible members: result/evidence/report/goods/effect. Examples: produce results; produce evidence; produce goods.
Sense check: determine which sense of produce is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with produce. Then verify with corpus evidence rather than intuition alone.
24. create
Illustrative semantic class: brought-into-being states or entities. Possible members: opportunity/problem/system/impression/value. Examples: create opportunities; create a problem; create value.
Sense check: determine which sense of create is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with create. Then verify with corpus evidence rather than intuition alone.
25. cause
Illustrative semantic class: consequences, often undesirable or consequential. Possible members: damage/problem/delay/confusion/change. Examples: cause damage; cause delays; cause confusion.
Sense check: determine which sense of cause is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with cause. Then verify with corpus evidence rather than intuition alone.
26. trigger
Illustrative semantic class: events or reactions initiated suddenly. Possible members: response/reaction/event/process/alarm. Examples: trigger a response; trigger an alarm; trigger a reaction.
Sense check: determine which sense of trigger is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with trigger. Then verify with corpus evidence rather than intuition alone.
27. prompt
Illustrative semantic class: responses or reconsideration. Possible members: question/review/action/debate/response. Examples: prompt a review; prompt action; prompt debate.
Sense check: determine which sense of prompt is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with prompt. Then verify with corpus evidence rather than intuition alone.
28. alleviate
Illustrative semantic class: undesirable states reduced. Possible members: pain/pressure/burden/problem/shortage. Examples: alleviate pain; alleviate pressure; alleviate shortages.
Sense check: determine which sense of alleviate is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with alleviate. Then verify with corpus evidence rather than intuition alone.
29. reduce
Illustrative semantic class: quantities or undesirable levels. Possible members: cost/risk/pressure/emission/error. Examples: reduce costs; reduce risk; reduce emissions.
Sense check: determine which sense of reduce is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with reduce. Then verify with corpus evidence rather than intuition alone.
30. increase
Illustrative semantic class: quantities, probabilities or levels. Possible members: risk/chance/cost/pressure/awareness. Examples: increase risk; increase awareness; increase costs.
Sense check: determine which sense of increase is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with increase. Then verify with corpus evidence rather than intuition alone.
31. enhance
Illustrative semantic class: desirable qualities or capacities. Possible members: quality/performance/understanding/value/security. Examples: enhance performance; enhance understanding; enhance security.
Sense check: determine which sense of enhance is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with enhance. Then verify with corpus evidence rather than intuition alone.
32. improve
Illustrative semantic class: states evaluated as better. Possible members: quality/performance/access/efficiency/health. Examples: improve efficiency; improve access; improve health.
Sense check: determine which sense of improve is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with improve. Then verify with corpus evidence rather than intuition alone.
33. undermine
Illustrative semantic class: valued structures or states weakened. Possible members: confidence/authority/stability/effort/credibility. Examples: undermine confidence; undermine stability; undermine credibility.
Sense check: determine which sense of undermine is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with undermine. Then verify with corpus evidence rather than intuition alone.
34. strengthen
Illustrative semantic class: structures or capacities made stronger. Possible members: relationship/system/evidence/control/capacity. Examples: strengthen relationships; strengthen control; strengthen capacity.
Sense check: determine which sense of strengthen is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with strengthen. Then verify with corpus evidence rather than intuition alone.
35. support
Illustrative semantic class: propositions, actions or entities receiving backing. Possible members: claim/argument/decision/development/person. Examples: support a claim; support development; support a decision.
Sense check: determine which sense of support is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with support. Then verify with corpus evidence rather than intuition alone.
36. challenge
Illustrative semantic class: positions, beliefs or authority contested. Possible members: assumption/decision/view/authority/claim. Examples: challenge an assumption; challenge a claim; challenge authority.
Sense check: determine which sense of challenge is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with challenge. Then verify with corpus evidence rather than intuition alone.
37. reject
Illustrative semantic class: propositions or formal submissions refused. Possible members: claim/proposal/request/idea/application. Examples: reject a proposal; reject a claim; reject an application.
Sense check: determine which sense of reject is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with reject. Then verify with corpus evidence rather than intuition alone.
38. accept
Illustrative semantic class: commitments, proposals or states acknowledged. Possible members: responsibility/offer/proposal/payment/reality. Examples: accept responsibility; accept an offer; accept reality.
Sense check: determine which sense of accept is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with accept. Then verify with corpus evidence rather than intuition alone.
39. express
Illustrative semantic class: attitudes or communicative states. Possible members: concern/interest/opinion/support/gratitude. Examples: express concern; express interest; express gratitude.
Sense check: determine which sense of express is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with express. Then verify with corpus evidence rather than intuition alone.
40. voice
Illustrative semantic class: publicly expressed attitudes. Possible members: concern/opposition/support/opinion/frustration. Examples: voice concerns; voice opposition; voice support.
Sense check: determine which sense of voice is active. A different sense may have a different semantic preference and should not be merged automatically.
Register check: compare news, academic, conversation and specialist corpora. The preferred semantic class may strengthen, weaken or change across registers.
Grammar check: identify whether the preference occurs in object, subject, modifier or complement position. Semantic preference is tied to co-selection in structure.
Production test: give a new noun from the same class and ask whether it combines naturally with voice. Then verify with corpus evidence rather than intuition alone.
25. Semantic-preference casebook — Cases 41–60
41. demonstrate
Illustrative semantic class: qualities, effects or propositions shown. Possible members: ability/effect/commitment/value/relationship. Examples: demonstrate ability; demonstrate commitment; demonstrate an effect.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace demonstrate with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when demonstrate appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by demonstrate.
42. indicate
Illustrative semantic class: evidence-linked propositions or states. Possible members: trend/result/problem/possibility/difference. Examples: indicate a trend; indicate a problem; indicate a difference.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace indicate with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when indicate appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by indicate.
43. suggest
Illustrative semantic class: propositions, interpretations or proposals. Possible members: possibility/result/relationship/approach/idea. Examples: suggest a relationship; suggest an approach; suggest a possibility.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace suggest with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when suggest appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by suggest.
44. reveal
Illustrative semantic class: previously hidden information. Possible members: pattern/problem/difference/identity/detail. Examples: reveal a pattern; reveal a problem; reveal details.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace reveal with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when reveal appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by reveal.
45. identify
Illustrative semantic class: diagnosed entities or features. Possible members: problem/factor/risk/pattern/need. Examples: identify a problem; identify risk; identify a pattern.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace identify with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when identify appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by identify.
46. assess
Illustrative semantic class: things evaluated systematically. Possible members: risk/performance/impact/need/quality. Examples: assess risk; assess performance; assess impact.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace assess with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when assess appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by assess.
47. evaluate
Illustrative semantic class: objects of systematic judgement. Possible members: evidence/performance/programme/option/effect. Examples: evaluate evidence; evaluate options; evaluate effects.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace evaluate with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when evaluate appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by evaluate.
48. measure
Illustrative semantic class: quantifiable properties or outcomes. Possible members: temperature/performance/effect/level/progress. Examples: measure progress; measure temperature; measure effects.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace measure with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when measure appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by measure.
49. monitor
Illustrative semantic class: ongoing states or processes watched. Possible members: progress/performance/condition/activity/risk. Examples: monitor progress; monitor risk; monitor activity.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace monitor with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when monitor appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by monitor.
50. manage
Illustrative semantic class: complex entities requiring control. Possible members: risk/project/team/resource/problem. Examples: manage risk; manage a project; manage resources.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace manage with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when manage appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by manage.
51. allocate
Illustrative semantic class: distributable resources or responsibilities. Possible members: resource/fund/time/budget/task. Examples: allocate resources; allocate funds; allocate time.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace allocate with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when allocate appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by allocate.
52. consume
Illustrative semantic class: resources or consumable inputs. Possible members: energy/resource/food/time/content. Examples: consume energy; consume resources; consume content.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace consume with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when consume appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by consume.
53. emit
Illustrative semantic class: outputs released from a source. Possible members: light/sound/gas/radiation/signal. Examples: emit light; emit gas; emit a signal.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace emit with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when emit appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by emit.
54. absorb
Illustrative semantic class: things taken in physically or metaphorically. Possible members: water/energy/cost/information/shock. Examples: absorb water; absorb costs; absorb information.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace absorb with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when absorb appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by absorb.
55. withstand
Illustrative semantic class: stressors or challenges resisted. Possible members: pressure/force/heat/criticism/test. Examples: withstand pressure; withstand heat; withstand criticism.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace withstand with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when withstand appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by withstand.
56. undergo
Illustrative semantic class: processes experienced by a participant. Possible members: treatment/surgery/change/transformation/process. Examples: undergo treatment; undergo change; undergo a transformation.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace undergo with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when undergo appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by undergo.
57. experience
Illustrative semantic class: events, conditions or transitions lived through. Possible members: change/growth/problem/difficulty/increase. Examples: experience growth; experience difficulties; experience change.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace experience with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when experience appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by experience.
58. face
Illustrative semantic class: difficulties or consequential situations confronted. Possible members: challenge/problem/risk/pressure/decision. Examples: face a challenge; face pressure; face a decision.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace face with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when face appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by face.
59. encounter
Illustrative semantic class: entities or difficulties met. Possible members: problem/difficulty/resistance/person/obstacle. Examples: encounter problems; encounter resistance; encounter an obstacle.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace encounter with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when encounter appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by encounter.
60. overcome
Illustrative semantic class: hindrances successfully passed. Possible members: difficulty/obstacle/problem/fear/barrier. Examples: overcome obstacles; overcome fear; overcome barriers.
Translation test: compare a translation equivalent in another language. Does the corresponding word select the same semantic class, a narrower class or a different construction?
Near-synonym test: replace overcome with a near-synonym. If the same semantic class no longer sounds natural, semantic preference helps explain why dictionary overlap does not guarantee interchangeability.
Comprehension test: when overcome appears, predict what kind of meaning is likely to follow before reading the next word. This turns semantic preference into a processing expectation.
Writing test: inspect a learner sentence that is grammatical but odd. Ask whether the noun belongs to a semantic class normally selected by overcome.
26. Learning semantic preference from corpora
1. address
Step 1 — collect: gather authentic concordance lines containing address. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as issues/problems/concerns/questions by meaning. Test whether the class problem or topic entities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
2. pose
Step 1 — collect: gather authentic concordance lines containing pose. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as risk/threat/challenge/problem/question by meaning. Test whether the class potential difficulty or discourse challenge explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
3. raise
Step 1 — collect: gather authentic concordance lines containing raise. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as question/issue/concern/awareness by meaning. Test whether the class discourse or public-attention items explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
4. meet
Step 1 — collect: gather authentic concordance lines containing meet. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as need/demand/requirement/target/deadline by meaning. Test whether the class standards, needs or goals explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
5. reach
Step 1 — collect: gather authentic concordance lines containing reach. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as agreement/conclusion/decision/target/destination by meaning. Test whether the class endpoints or achieved states explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
6. draw
Step 1 — collect: gather authentic concordance lines containing draw. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as conclusion/distinction/comparison/attention by meaning. Test whether the class cognitive or discourse operations explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
7. conduct
Step 1 — collect: gather authentic concordance lines containing conduct. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as research/study/survey/investigation/interview by meaning. Test whether the class formal organised activities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
8. perform
Step 1 — collect: gather authentic concordance lines containing perform. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as task/test/function/operation/procedure by meaning. Test whether the class deliberate activities or functions explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
9. provide
Step 1 — collect: gather authentic concordance lines containing provide. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as information/support/service/evidence/access by meaning. Test whether the class resources or enabling inputs explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
10. offer
Step 1 — collect: gather authentic concordance lines containing offer. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as help/advice/service/opportunity/choice by meaning. Test whether the class benefits, resources or possibilities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
11. gain
Step 1 — collect: gather authentic concordance lines containing gain. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as experience/access/insight/confidence/support by meaning. Test whether the class acquired resources or positive states explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
12. acquire
Step 1 — collect: gather authentic concordance lines containing acquire. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as knowledge/skill/language/property/company by meaning. Test whether the class things obtained or learned explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
13. develop
Step 1 — collect: gather authentic concordance lines containing develop. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as skill/system/method/strategy/disease by meaning. Test whether the class emerging capabilities, systems or conditions explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
14. establish
Step 1 — collect: gather authentic concordance lines containing establish. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as relationship/system/fact/principle/company by meaning. Test whether the class entities, relations or propositions made stable explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
15. maintain
Step 1 — collect: gather authentic concordance lines containing maintain. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as standard/system/relationship/pressure/balance by meaning. Test whether the class states or arrangements kept stable explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
16. achieve
Step 1 — collect: gather authentic concordance lines containing achieve. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as goal/result/outcome/success/balance by meaning. Test whether the class desired endpoints or states explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
17. obtain
Step 1 — collect: gather authentic concordance lines containing obtain. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as permission/information/result/access/evidence by meaning. Test whether the class formally acquired resources or outcomes explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
18. submit
Step 1 — collect: gather authentic concordance lines containing submit. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as application/report/request/proposal/evidence by meaning. Test whether the class formal documents or claims explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
19. issue
Step 1 — collect: gather authentic concordance lines containing issue. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as warning/statement/report/guidance/order by meaning. Test whether the class official communicative documents/actions explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
20. launch
Step 1 — collect: gather authentic concordance lines containing launch. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as campaign/product/service/inquiry/programme by meaning. Test whether the class initiated public or organised activities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
21. deliver
Step 1 — collect: gather authentic concordance lines containing deliver. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as service/result/speech/message/package by meaning. Test whether the class outputs, performances or transferred items explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
22. generate
Step 1 — collect: gather authentic concordance lines containing generate. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as data/revenue/interest/electricity/idea by meaning. Test whether the class produced outputs explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
23. produce
Step 1 — collect: gather authentic concordance lines containing produce. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as result/evidence/report/goods/effect by meaning. Test whether the class created outputs explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
24. create
Step 1 — collect: gather authentic concordance lines containing create. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as opportunity/problem/system/impression/value by meaning. Test whether the class brought-into-being states or entities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
25. cause
Step 1 — collect: gather authentic concordance lines containing cause. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as damage/problem/delay/confusion/change by meaning. Test whether the class consequences, often undesirable or consequential explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
26. trigger
Step 1 — collect: gather authentic concordance lines containing trigger. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as response/reaction/event/process/alarm by meaning. Test whether the class events or reactions initiated suddenly explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
27. prompt
Step 1 — collect: gather authentic concordance lines containing prompt. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as question/review/action/debate/response by meaning. Test whether the class responses or reconsideration explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
28. alleviate
Step 1 — collect: gather authentic concordance lines containing alleviate. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as pain/pressure/burden/problem/shortage by meaning. Test whether the class undesirable states reduced explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
29. reduce
Step 1 — collect: gather authentic concordance lines containing reduce. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as cost/risk/pressure/emission/error by meaning. Test whether the class quantities or undesirable levels explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
30. increase
Step 1 — collect: gather authentic concordance lines containing increase. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as risk/chance/cost/pressure/awareness by meaning. Test whether the class quantities, probabilities or levels explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
31. enhance
Step 1 — collect: gather authentic concordance lines containing enhance. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as quality/performance/understanding/value/security by meaning. Test whether the class desirable qualities or capacities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
32. improve
Step 1 — collect: gather authentic concordance lines containing improve. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as quality/performance/access/efficiency/health by meaning. Test whether the class states evaluated as better explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
33. undermine
Step 1 — collect: gather authentic concordance lines containing undermine. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as confidence/authority/stability/effort/credibility by meaning. Test whether the class valued structures or states weakened explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
34. strengthen
Step 1 — collect: gather authentic concordance lines containing strengthen. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as relationship/system/evidence/control/capacity by meaning. Test whether the class structures or capacities made stronger explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
35. support
Step 1 — collect: gather authentic concordance lines containing support. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as claim/argument/decision/development/person by meaning. Test whether the class propositions, actions or entities receiving backing explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
36. challenge
Step 1 — collect: gather authentic concordance lines containing challenge. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as assumption/decision/view/authority/claim by meaning. Test whether the class positions, beliefs or authority contested explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
37. reject
Step 1 — collect: gather authentic concordance lines containing reject. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as claim/proposal/request/idea/application by meaning. Test whether the class propositions or formal submissions refused explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
38. accept
Step 1 — collect: gather authentic concordance lines containing accept. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as responsibility/offer/proposal/payment/reality by meaning. Test whether the class commitments, proposals or states acknowledged explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
39. express
Step 1 — collect: gather authentic concordance lines containing express. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as concern/interest/opinion/support/gratitude by meaning. Test whether the class attitudes or communicative states explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
40. voice
Step 1 — collect: gather authentic concordance lines containing voice. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as concern/opposition/support/opinion/frustration by meaning. Test whether the class publicly expressed attitudes explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
41. demonstrate
Step 1 — collect: gather authentic concordance lines containing demonstrate. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as ability/effect/commitment/value/relationship by meaning. Test whether the class qualities, effects or propositions shown explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
42. indicate
Step 1 — collect: gather authentic concordance lines containing indicate. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as trend/result/problem/possibility/difference by meaning. Test whether the class evidence-linked propositions or states explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
43. suggest
Step 1 — collect: gather authentic concordance lines containing suggest. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as possibility/result/relationship/approach/idea by meaning. Test whether the class propositions, interpretations or proposals explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
44. reveal
Step 1 — collect: gather authentic concordance lines containing reveal. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as pattern/problem/difference/identity/detail by meaning. Test whether the class previously hidden information explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
45. identify
Step 1 — collect: gather authentic concordance lines containing identify. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as problem/factor/risk/pattern/need by meaning. Test whether the class diagnosed entities or features explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
46. assess
Step 1 — collect: gather authentic concordance lines containing assess. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as risk/performance/impact/need/quality by meaning. Test whether the class things evaluated systematically explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
47. evaluate
Step 1 — collect: gather authentic concordance lines containing evaluate. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as evidence/performance/programme/option/effect by meaning. Test whether the class objects of systematic judgement explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
48. measure
Step 1 — collect: gather authentic concordance lines containing measure. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as temperature/performance/effect/level/progress by meaning. Test whether the class quantifiable properties or outcomes explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
49. monitor
Step 1 — collect: gather authentic concordance lines containing monitor. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as progress/performance/condition/activity/risk by meaning. Test whether the class ongoing states or processes watched explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
50. manage
Step 1 — collect: gather authentic concordance lines containing manage. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as risk/project/team/resource/problem by meaning. Test whether the class complex entities requiring control explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
51. allocate
Step 1 — collect: gather authentic concordance lines containing allocate. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as resource/fund/time/budget/task by meaning. Test whether the class distributable resources or responsibilities explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
52. consume
Step 1 — collect: gather authentic concordance lines containing consume. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as energy/resource/food/time/content by meaning. Test whether the class resources or consumable inputs explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
53. emit
Step 1 — collect: gather authentic concordance lines containing emit. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as light/sound/gas/radiation/signal by meaning. Test whether the class outputs released from a source explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
54. absorb
Step 1 — collect: gather authentic concordance lines containing absorb. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as water/energy/cost/information/shock by meaning. Test whether the class things taken in physically or metaphorically explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
55. withstand
Step 1 — collect: gather authentic concordance lines containing withstand. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as pressure/force/heat/criticism/test by meaning. Test whether the class stressors or challenges resisted explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
56. undergo
Step 1 — collect: gather authentic concordance lines containing undergo. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as treatment/surgery/change/transformation/process by meaning. Test whether the class processes experienced by a participant explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
57. experience
Step 1 — collect: gather authentic concordance lines containing experience. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as change/growth/problem/difficulty/increase by meaning. Test whether the class events, conditions or transitions lived through explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
58. face
Step 1 — collect: gather authentic concordance lines containing face. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as challenge/problem/risk/pressure/decision by meaning. Test whether the class difficulties or consequential situations confronted explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
59. encounter
Step 1 — collect: gather authentic concordance lines containing encounter. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as problem/difficulty/resistance/person/obstacle by meaning. Test whether the class entities or difficulties met explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
60. overcome
Step 1 — collect: gather authentic concordance lines containing overcome. Focus on the same sense and grammatical pattern before grouping examples.
Step 2 — cluster: group collocates such as difficulty/obstacle/problem/fear/barrier by meaning. Test whether the class hindrances successfully passed explains a meaningful portion of the data rather than being invented after the fact.
Step 3 — compare: inspect a near-synonym and see whether it selects the same class. Differences reveal lexical conventionality invisible in definitions.
Step 4 — produce: generate a new example using a different member of the same class, then verify naturalness with corpus or dictionary evidence.
27. Frequently asked questions
What is semantic preference?
A tendency for a word or phrase to occur with members of a recurring semantic class.
How is semantic preference different from collocation?
Collocation links particular words; semantic preference groups several collocates by shared meaning.
How is it different from semantic prosody?
Semantic prosody concerns evaluative colouring; semantic preference concerns the semantic category of neighbours.
Is semantic preference a grammar rule?
No. It is a probabilistic usage tendency.
How is semantic preference found?
By examining many corpus examples and clustering recurrent collocates by meaning.
Can one word have several preferences?
Yes, especially across different senses, grammatical positions or registers.
Why does it matter for learners?
It helps explain natural word choice beyond dictionary definitions.
Can semantic preference help reading?
Yes. Recurrent patterns create expectations about what kinds of meanings are likely to follow.
Can translation equivalents differ?
Yes. Similar denotation does not guarantee the same semantic preferences across languages.
What is the simplest rule?
Look past one collocate and ask what meaning class the word repeatedly selects.
28. Research grounding
Cambridge corpus-linguistics work treats semantic preference as part of larger units of meaning alongside collocation and colligation. Recent Cambridge material distinguishes semantic preference from semantic prosody and notes that both can show register associations. Lexical-priming theory likewise proposes that repeated language experience primes words for collocation, semantic association, colligation and pragmatic functions.
- Cambridge — Semantic preference, semantic prosody and lexical priming
- Cambridge — Unit of meaning, collocation, colligation and semantic preference
29. eduKateSG routes
- Collocations and Word Partnerships
- Semantic Prosody
- Lexical Priming
- Corpus Linguistics
- Vocabulary Master Router
30. Final model
Semantic preference reveals a layer of lexical knowledge between individual collocation and broad meaning. Words do not merely prefer particular partners; they can repeatedly select whole classes of meaning.
Learners who notice those classes gain a stronger sense of what a word naturally does in real text and why some grammatically possible combinations remain unusual.
Know the word’s partners. Then learn the kind of company it usually keeps.
31. Final semantic-preference audit
1. address
Class-boundary test: propose one new noun that clearly belongs to problem or topic entities and one that does not. Predict which should combine more naturally with address, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach address with several class members—issues/problems/concerns/questions—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
2. pose
Class-boundary test: propose one new noun that clearly belongs to potential difficulty or discourse challenge and one that does not. Predict which should combine more naturally with pose, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach pose with several class members—risk/threat/challenge/problem/question—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
3. raise
Class-boundary test: propose one new noun that clearly belongs to discourse or public-attention items and one that does not. Predict which should combine more naturally with raise, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach raise with several class members—question/issue/concern/awareness—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
4. meet
Class-boundary test: propose one new noun that clearly belongs to standards, needs or goals and one that does not. Predict which should combine more naturally with meet, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach meet with several class members—need/demand/requirement/target/deadline—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
5. reach
Class-boundary test: propose one new noun that clearly belongs to endpoints or achieved states and one that does not. Predict which should combine more naturally with reach, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach reach with several class members—agreement/conclusion/decision/target/destination—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
6. draw
Class-boundary test: propose one new noun that clearly belongs to cognitive or discourse operations and one that does not. Predict which should combine more naturally with draw, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach draw with several class members—conclusion/distinction/comparison/attention—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
7. conduct
Class-boundary test: propose one new noun that clearly belongs to formal organised activities and one that does not. Predict which should combine more naturally with conduct, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach conduct with several class members—research/study/survey/investigation/interview—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
8. perform
Class-boundary test: propose one new noun that clearly belongs to deliberate activities or functions and one that does not. Predict which should combine more naturally with perform, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach perform with several class members—task/test/function/operation/procedure—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
9. provide
Class-boundary test: propose one new noun that clearly belongs to resources or enabling inputs and one that does not. Predict which should combine more naturally with provide, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach provide with several class members—information/support/service/evidence/access—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
10. offer
Class-boundary test: propose one new noun that clearly belongs to benefits, resources or possibilities and one that does not. Predict which should combine more naturally with offer, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach offer with several class members—help/advice/service/opportunity/choice—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
11. gain
Class-boundary test: propose one new noun that clearly belongs to acquired resources or positive states and one that does not. Predict which should combine more naturally with gain, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach gain with several class members—experience/access/insight/confidence/support—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
12. acquire
Class-boundary test: propose one new noun that clearly belongs to things obtained or learned and one that does not. Predict which should combine more naturally with acquire, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach acquire with several class members—knowledge/skill/language/property/company—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
13. develop
Class-boundary test: propose one new noun that clearly belongs to emerging capabilities, systems or conditions and one that does not. Predict which should combine more naturally with develop, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach develop with several class members—skill/system/method/strategy/disease—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
14. establish
Class-boundary test: propose one new noun that clearly belongs to entities, relations or propositions made stable and one that does not. Predict which should combine more naturally with establish, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach establish with several class members—relationship/system/fact/principle/company—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
15. maintain
Class-boundary test: propose one new noun that clearly belongs to states or arrangements kept stable and one that does not. Predict which should combine more naturally with maintain, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach maintain with several class members—standard/system/relationship/pressure/balance—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
16. achieve
Class-boundary test: propose one new noun that clearly belongs to desired endpoints or states and one that does not. Predict which should combine more naturally with achieve, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach achieve with several class members—goal/result/outcome/success/balance—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
17. obtain
Class-boundary test: propose one new noun that clearly belongs to formally acquired resources or outcomes and one that does not. Predict which should combine more naturally with obtain, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach obtain with several class members—permission/information/result/access/evidence—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
18. submit
Class-boundary test: propose one new noun that clearly belongs to formal documents or claims and one that does not. Predict which should combine more naturally with submit, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach submit with several class members—application/report/request/proposal/evidence—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
19. issue
Class-boundary test: propose one new noun that clearly belongs to official communicative documents/actions and one that does not. Predict which should combine more naturally with issue, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach issue with several class members—warning/statement/report/guidance/order—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
20. launch
Class-boundary test: propose one new noun that clearly belongs to initiated public or organised activities and one that does not. Predict which should combine more naturally with launch, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach launch with several class members—campaign/product/service/inquiry/programme—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
21. deliver
Class-boundary test: propose one new noun that clearly belongs to outputs, performances or transferred items and one that does not. Predict which should combine more naturally with deliver, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach deliver with several class members—service/result/speech/message/package—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
22. generate
Class-boundary test: propose one new noun that clearly belongs to produced outputs and one that does not. Predict which should combine more naturally with generate, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach generate with several class members—data/revenue/interest/electricity/idea—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
23. produce
Class-boundary test: propose one new noun that clearly belongs to created outputs and one that does not. Predict which should combine more naturally with produce, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach produce with several class members—result/evidence/report/goods/effect—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
24. create
Class-boundary test: propose one new noun that clearly belongs to brought-into-being states or entities and one that does not. Predict which should combine more naturally with create, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach create with several class members—opportunity/problem/system/impression/value—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
25. cause
Class-boundary test: propose one new noun that clearly belongs to consequences, often undesirable or consequential and one that does not. Predict which should combine more naturally with cause, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach cause with several class members—damage/problem/delay/confusion/change—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
26. trigger
Class-boundary test: propose one new noun that clearly belongs to events or reactions initiated suddenly and one that does not. Predict which should combine more naturally with trigger, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach trigger with several class members—response/reaction/event/process/alarm—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
27. prompt
Class-boundary test: propose one new noun that clearly belongs to responses or reconsideration and one that does not. Predict which should combine more naturally with prompt, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach prompt with several class members—question/review/action/debate/response—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
28. alleviate
Class-boundary test: propose one new noun that clearly belongs to undesirable states reduced and one that does not. Predict which should combine more naturally with alleviate, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach alleviate with several class members—pain/pressure/burden/problem/shortage—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
29. reduce
Class-boundary test: propose one new noun that clearly belongs to quantities or undesirable levels and one that does not. Predict which should combine more naturally with reduce, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach reduce with several class members—cost/risk/pressure/emission/error—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
30. increase
Class-boundary test: propose one new noun that clearly belongs to quantities, probabilities or levels and one that does not. Predict which should combine more naturally with increase, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach increase with several class members—risk/chance/cost/pressure/awareness—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
31. enhance
Class-boundary test: propose one new noun that clearly belongs to desirable qualities or capacities and one that does not. Predict which should combine more naturally with enhance, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach enhance with several class members—quality/performance/understanding/value/security—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
32. improve
Class-boundary test: propose one new noun that clearly belongs to states evaluated as better and one that does not. Predict which should combine more naturally with improve, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach improve with several class members—quality/performance/access/efficiency/health—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
33. undermine
Class-boundary test: propose one new noun that clearly belongs to valued structures or states weakened and one that does not. Predict which should combine more naturally with undermine, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach undermine with several class members—confidence/authority/stability/effort/credibility—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
34. strengthen
Class-boundary test: propose one new noun that clearly belongs to structures or capacities made stronger and one that does not. Predict which should combine more naturally with strengthen, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach strengthen with several class members—relationship/system/evidence/control/capacity—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
35. support
Class-boundary test: propose one new noun that clearly belongs to propositions, actions or entities receiving backing and one that does not. Predict which should combine more naturally with support, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach support with several class members—claim/argument/decision/development/person—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
36. challenge
Class-boundary test: propose one new noun that clearly belongs to positions, beliefs or authority contested and one that does not. Predict which should combine more naturally with challenge, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach challenge with several class members—assumption/decision/view/authority/claim—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
37. reject
Class-boundary test: propose one new noun that clearly belongs to propositions or formal submissions refused and one that does not. Predict which should combine more naturally with reject, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach reject with several class members—claim/proposal/request/idea/application—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
38. accept
Class-boundary test: propose one new noun that clearly belongs to commitments, proposals or states acknowledged and one that does not. Predict which should combine more naturally with accept, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach accept with several class members—responsibility/offer/proposal/payment/reality—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
39. express
Class-boundary test: propose one new noun that clearly belongs to attitudes or communicative states and one that does not. Predict which should combine more naturally with express, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach express with several class members—concern/interest/opinion/support/gratitude—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
40. voice
Class-boundary test: propose one new noun that clearly belongs to publicly expressed attitudes and one that does not. Predict which should combine more naturally with voice, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach voice with several class members—concern/opposition/support/opinion/frustration—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
41. demonstrate
Class-boundary test: propose one new noun that clearly belongs to qualities, effects or propositions shown and one that does not. Predict which should combine more naturally with demonstrate, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach demonstrate with several class members—ability/effect/commitment/value/relationship—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
42. indicate
Class-boundary test: propose one new noun that clearly belongs to evidence-linked propositions or states and one that does not. Predict which should combine more naturally with indicate, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach indicate with several class members—trend/result/problem/possibility/difference—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
43. suggest
Class-boundary test: propose one new noun that clearly belongs to propositions, interpretations or proposals and one that does not. Predict which should combine more naturally with suggest, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach suggest with several class members—possibility/result/relationship/approach/idea—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
44. reveal
Class-boundary test: propose one new noun that clearly belongs to previously hidden information and one that does not. Predict which should combine more naturally with reveal, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach reveal with several class members—pattern/problem/difference/identity/detail—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
45. identify
Class-boundary test: propose one new noun that clearly belongs to diagnosed entities or features and one that does not. Predict which should combine more naturally with identify, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach identify with several class members—problem/factor/risk/pattern/need—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
46. assess
Class-boundary test: propose one new noun that clearly belongs to things evaluated systematically and one that does not. Predict which should combine more naturally with assess, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach assess with several class members—risk/performance/impact/need/quality—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
47. evaluate
Class-boundary test: propose one new noun that clearly belongs to objects of systematic judgement and one that does not. Predict which should combine more naturally with evaluate, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach evaluate with several class members—evidence/performance/programme/option/effect—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
48. measure
Class-boundary test: propose one new noun that clearly belongs to quantifiable properties or outcomes and one that does not. Predict which should combine more naturally with measure, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach measure with several class members—temperature/performance/effect/level/progress—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
49. monitor
Class-boundary test: propose one new noun that clearly belongs to ongoing states or processes watched and one that does not. Predict which should combine more naturally with monitor, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach monitor with several class members—progress/performance/condition/activity/risk—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
50. manage
Class-boundary test: propose one new noun that clearly belongs to complex entities requiring control and one that does not. Predict which should combine more naturally with manage, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach manage with several class members—risk/project/team/resource/problem—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
51. allocate
Class-boundary test: propose one new noun that clearly belongs to distributable resources or responsibilities and one that does not. Predict which should combine more naturally with allocate, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach allocate with several class members—resource/fund/time/budget/task—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
52. consume
Class-boundary test: propose one new noun that clearly belongs to resources or consumable inputs and one that does not. Predict which should combine more naturally with consume, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach consume with several class members—energy/resource/food/time/content—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
53. emit
Class-boundary test: propose one new noun that clearly belongs to outputs released from a source and one that does not. Predict which should combine more naturally with emit, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach emit with several class members—light/sound/gas/radiation/signal—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
54. absorb
Class-boundary test: propose one new noun that clearly belongs to things taken in physically or metaphorically and one that does not. Predict which should combine more naturally with absorb, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach absorb with several class members—water/energy/cost/information/shock—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
55. withstand
Class-boundary test: propose one new noun that clearly belongs to stressors or challenges resisted and one that does not. Predict which should combine more naturally with withstand, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach withstand with several class members—pressure/force/heat/criticism/test—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
56. undergo
Class-boundary test: propose one new noun that clearly belongs to processes experienced by a participant and one that does not. Predict which should combine more naturally with undergo, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach undergo with several class members—treatment/surgery/change/transformation/process—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
57. experience
Class-boundary test: propose one new noun that clearly belongs to events, conditions or transitions lived through and one that does not. Predict which should combine more naturally with experience, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach experience with several class members—change/growth/problem/difficulty/increase—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
58. face
Class-boundary test: propose one new noun that clearly belongs to difficulties or consequential situations confronted and one that does not. Predict which should combine more naturally with face, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach face with several class members—challenge/problem/risk/pressure/decision—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
59. encounter
Class-boundary test: propose one new noun that clearly belongs to entities or difficulties met and one that does not. Predict which should combine more naturally with encounter, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach encounter with several class members—problem/difficulty/resistance/person/obstacle—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
60. overcome
Class-boundary test: propose one new noun that clearly belongs to hindrances successfully passed and one that does not. Predict which should combine more naturally with overcome, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach overcome with several class members—difficulty/obstacle/problem/fear/barrier—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
61. address
Class-boundary test: propose one new noun that clearly belongs to problem or topic entities and one that does not. Predict which should combine more naturally with address, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach address with several class members—issues/problems/concerns/questions—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
62. pose
Class-boundary test: propose one new noun that clearly belongs to potential difficulty or discourse challenge and one that does not. Predict which should combine more naturally with pose, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach pose with several class members—risk/threat/challenge/problem/question—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
63. raise
Class-boundary test: propose one new noun that clearly belongs to discourse or public-attention items and one that does not. Predict which should combine more naturally with raise, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach raise with several class members—question/issue/concern/awareness—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
64. meet
Class-boundary test: propose one new noun that clearly belongs to standards, needs or goals and one that does not. Predict which should combine more naturally with meet, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach meet with several class members—need/demand/requirement/target/deadline—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
65. reach
Class-boundary test: propose one new noun that clearly belongs to endpoints or achieved states and one that does not. Predict which should combine more naturally with reach, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach reach with several class members—agreement/conclusion/decision/target/destination—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
66. draw
Class-boundary test: propose one new noun that clearly belongs to cognitive or discourse operations and one that does not. Predict which should combine more naturally with draw, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach draw with several class members—conclusion/distinction/comparison/attention—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
67. conduct
Class-boundary test: propose one new noun that clearly belongs to formal organised activities and one that does not. Predict which should combine more naturally with conduct, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach conduct with several class members—research/study/survey/investigation/interview—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
68. perform
Class-boundary test: propose one new noun that clearly belongs to deliberate activities or functions and one that does not. Predict which should combine more naturally with perform, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach perform with several class members—task/test/function/operation/procedure—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
69. provide
Class-boundary test: propose one new noun that clearly belongs to resources or enabling inputs and one that does not. Predict which should combine more naturally with provide, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach provide with several class members—information/support/service/evidence/access—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
70. offer
Class-boundary test: propose one new noun that clearly belongs to benefits, resources or possibilities and one that does not. Predict which should combine more naturally with offer, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach offer with several class members—help/advice/service/opportunity/choice—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
71. gain
Class-boundary test: propose one new noun that clearly belongs to acquired resources or positive states and one that does not. Predict which should combine more naturally with gain, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach gain with several class members—experience/access/insight/confidence/support—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
72. acquire
Class-boundary test: propose one new noun that clearly belongs to things obtained or learned and one that does not. Predict which should combine more naturally with acquire, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach acquire with several class members—knowledge/skill/language/property/company—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
73. develop
Class-boundary test: propose one new noun that clearly belongs to emerging capabilities, systems or conditions and one that does not. Predict which should combine more naturally with develop, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach develop with several class members—skill/system/method/strategy/disease—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.
74. establish
Class-boundary test: propose one new noun that clearly belongs to entities, relations or propositions made stable and one that does not. Predict which should combine more naturally with establish, then verify the prediction in authentic data.
Preference-versus-rule check: find one legitimate example outside the dominant class. Explain why that exception does not invalidate a probabilistic preference.
Pedagogical summary: teach establish with several class members—relationship/system/fact/principle/company—so the learner abstracts a usage tendency rather than memorising only one fixed collocation.