How to Improve Vocabulary | Approximation and Precision — Use About, Roughly, Nearly, Almost, Up To, At Least and At Most Precisely

This specialist guide continues the How to Improve Vocabulary series by isolating approximation and precision as a word-choice problem. The purpose is not to collect impressive expressions. It is to make small differences in function visible enough that a learner can choose deliberately, explain the choice, and transfer the distinction into new reading, speaking and writing situations.

Approximation is not the opposite of precision. A careful estimate can be more truthful than a false exact number, while lower and upper bounds can be exact about what is known. The wider complete vocabulary system explains how words are learned and retained. This page is narrower: it owns the precision decisions named in the title and treats neighbouring systems only when they change those decisions.

The same discipline appears throughout the precision strand. Cause and Effect separates reasons from results; Causal Strength separates causal commitment from association; Purpose and Result separates intention from outcome; Comparison and Degree exposes scales and thresholds. Here the learner applies that style of reasoning to approximation and precision.

Strong vocabulary is quiet when it works. The reader does not stop to admire the connector or modifier; the reader simply receives the right relation. That means an ordinary word can be the expert choice. A formal word can be the wrong choice. The test is whether the expression preserves meaning, grammar, register and the writer’s actual evidence.

Quick Read: Approximate Honestly; Bound Explicitly

About and roughly give estimates near a value without claiming exactness; the acceptable tolerance depends on context.

Nearly and almost normally approach a target without reaching it in the intended reading.

Up to sets an upper bound but does not say that the maximum is typical or guaranteed.

At least sets a lower bound: the true value is no smaller than the stated amount.

At most sets an upper bound: the true value is no greater than the stated amount.

False precision can mislead. If the measurement is approximate, an unnecessarily exact-looking number may imply knowledge the writer does not have.

Scope Boundary

This owner covers approximate quantities, closeness to a target and lower or upper bounds. It does not own full numerical proportion or statistical uncertainty, though those systems may use the same expressions.

Use the Word-Choice Decision System whenever the difficulty is broader than this topic. Use Scope and Quantifiers for how large a claim becomes, Modality and Certainty for commitment to possibility or necessity, and Contrast and Concession when the main relation is opposition or resistant truth.

1. Core Principles

1. Match precision to measurement

Do not report more digits than the method can justify.

2. About is contextual

About ten people and about ten million dollars can involve very different tolerances.

3. Roughly signals estimation

It often suits early calculations, informal summaries and values where exactness is unnecessary.

4. Nearly approaches from below

In ordinary quantitative use, nearly ten suggests a value close to but less than ten.

5. Almost also marks near attainment

It can apply to quantities, completion, states and events.

6. Up to is a ceiling

It gives a maximum allowed, possible or observed value, not a typical value.

7. At least is a floor

It guarantees a minimum if the statement is accurate.

8. At most is a ceiling

It limits the maximum and may leave the lower range open.

9. Bounds can be more informative than guesses

When only a minimum or maximum is known, say so.

10. Marketing can exploit maxima

Up to can highlight the best case while typical outcomes are lower.

11. Context defines acceptable approximation

A minute may be negligible in one task and critical in another.

12. Precision includes uncertainty

Good writing states what is known at the resolution that evidence supports.

2. Core Forms

1. about

about has a specific job here. It presents a value as an estimate near the stated number. The nearest trap is exactly, which removes the approximation. Test it with this question: What tolerance would a reasonable reader infer in this context?

Model: About 200 people attended the open house. Near miss: Exactly about 200 people attended. Repair: About 200 people attended. Do not combine exactness and approximation without a special reason.

2. roughly

roughly has a specific job here. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. The nearest trap is approximately, which is often more formal but not universally more precise. Test it with this question: Is a coarse but useful estimate appropriate?

Model: The first stage took roughly twenty minutes. Near miss: The first stage took roughly 20.000 minutes. Repair: The first stage took roughly twenty minutes. Formatting can contradict the intended uncertainty.

3. nearly

nearly has a specific job here. It indicates that a value or state came close to a target without reaching it in the intended reading. The nearest trap is about, which can fall on either side of the estimate. Test it with this question: Is the value below but close to the reference point?

Model: Nearly 100 entries were submitted. Near miss: Nearly 100 entries were submitted, in fact there were 103. Repair: About 100 entries were submitted. When the count exceeds the target, nearly is usually wrong for that numerical reading.

4. almost

almost has a specific job here. It indicates near attainment, completion or occurrence without full attainment. The nearest trap is nearly, which overlaps strongly in many contexts. Test it with this question: Did the event or threshold remain just short of completion?

Model: The tank was almost full when the valve closed. Near miss: The tank was almost full and overflowing for an hour. Repair: The tank was full and overflowing for an hour. The surrounding facts must agree with non-attainment.

5. up to

up to has a specific job here. It sets an upper limit on a possible, allowed or observed value. The nearest trap is at least, which sets a lower rather than upper bound. Test it with this question: Am I stating a ceiling rather than a typical or guaranteed amount?

Model: The hall can hold up to 300 people. Near miss: The hall holds 300 people at every event because capacity is up to 300. Repair: The hall can hold up to 300 people. Maximum capacity is not typical attendance.

6. at least

at least has a specific job here. It sets a lower bound: the true or required value is no smaller than the stated amount. The nearest trap is at most, which reverses the bound. Test it with this question: Would any larger value still satisfy the statement?

Model: Bring at least two examples to the discussion. Near miss: Bring at least two examples, but no more than two. Repair: Bring exactly two examples. Combined bounds can collapse into an exact value.

7. at most

at most has a specific job here. It sets an upper bound: the true or permitted value is no greater than the stated amount. The nearest trap is up to, which often overlaps but may differ in register and pragmatic emphasis. Test it with this question: Would any smaller value still satisfy the statement?

Model: Each answer should contain at most three sentences. Near miss: Each answer should contain at most three sentences and must contain four. Repair: Each answer should contain at most three sentences. Check other rules for bound conflicts.

3. Pairwise Decisions

1. about versus roughly

about and roughly can occupy neighbouring territory, but they are not safe automatic replacements. It presents a value as an estimate near the stated number. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

2. about versus nearly

about and nearly can occupy neighbouring territory, but they are not safe automatic replacements. It presents a value as an estimate near the stated number. It indicates that a value or state came close to a target without reaching it in the intended reading. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

3. about versus almost

about and almost can occupy neighbouring territory, but they are not safe automatic replacements. It presents a value as an estimate near the stated number. It indicates near attainment, completion or occurrence without full attainment. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

4. about versus up to

about and up to can occupy neighbouring territory, but they are not safe automatic replacements. It presents a value as an estimate near the stated number. It sets an upper limit on a possible, allowed or observed value. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

5. about versus at least

about and at least can occupy neighbouring territory, but they are not safe automatic replacements. It presents a value as an estimate near the stated number. It sets a lower bound: the true or required value is no smaller than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

6. about versus at most

about and at most can occupy neighbouring territory, but they are not safe automatic replacements. It presents a value as an estimate near the stated number. It sets an upper bound: the true or permitted value is no greater than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

7. roughly versus nearly

roughly and nearly can occupy neighbouring territory, but they are not safe automatic replacements. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. It indicates that a value or state came close to a target without reaching it in the intended reading. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

8. roughly versus almost

roughly and almost can occupy neighbouring territory, but they are not safe automatic replacements. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. It indicates near attainment, completion or occurrence without full attainment. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

9. roughly versus up to

roughly and up to can occupy neighbouring territory, but they are not safe automatic replacements. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. It sets an upper limit on a possible, allowed or observed value. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

10. roughly versus at least

roughly and at least can occupy neighbouring territory, but they are not safe automatic replacements. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. It sets a lower bound: the true or required value is no smaller than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

11. roughly versus at most

roughly and at most can occupy neighbouring territory, but they are not safe automatic replacements. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. It sets an upper bound: the true or permitted value is no greater than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

12. nearly versus almost

nearly and almost can occupy neighbouring territory, but they are not safe automatic replacements. It indicates that a value or state came close to a target without reaching it in the intended reading. It indicates near attainment, completion or occurrence without full attainment. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

13. nearly versus up to

nearly and up to can occupy neighbouring territory, but they are not safe automatic replacements. It indicates that a value or state came close to a target without reaching it in the intended reading. It sets an upper limit on a possible, allowed or observed value. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

14. nearly versus at least

nearly and at least can occupy neighbouring territory, but they are not safe automatic replacements. It indicates that a value or state came close to a target without reaching it in the intended reading. It sets a lower bound: the true or required value is no smaller than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

15. nearly versus at most

nearly and at most can occupy neighbouring territory, but they are not safe automatic replacements. It indicates that a value or state came close to a target without reaching it in the intended reading. It sets an upper bound: the true or permitted value is no greater than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

16. almost versus up to

almost and up to can occupy neighbouring territory, but they are not safe automatic replacements. It indicates near attainment, completion or occurrence without full attainment. It sets an upper limit on a possible, allowed or observed value. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

17. almost versus at least

almost and at least can occupy neighbouring territory, but they are not safe automatic replacements. It indicates near attainment, completion or occurrence without full attainment. It sets a lower bound: the true or required value is no smaller than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

18. almost versus at most

almost and at most can occupy neighbouring territory, but they are not safe automatic replacements. It indicates near attainment, completion or occurrence without full attainment. It sets an upper bound: the true or permitted value is no greater than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

19. up to versus at least

up to and at least can occupy neighbouring territory, but they are not safe automatic replacements. It sets an upper limit on a possible, allowed or observed value. It sets a lower bound: the true or required value is no smaller than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

20. up to versus at most

up to and at most can occupy neighbouring territory, but they are not safe automatic replacements. It sets an upper limit on a possible, allowed or observed value. It sets an upper bound: the true or permitted value is no greater than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

21. at least versus at most

at least and at most can occupy neighbouring territory, but they are not safe automatic replacements. It sets a lower bound: the true or required value is no smaller than the stated amount. It sets an upper bound: the true or permitted value is no greater than the stated amount. To choose, identify the writer’s intended relation, the grammar that follows, the strength of the claim and the likely reader inference. Substitution is useful as a test precisely because a good substitution exposes what the original expression was doing.

4. Precision Laboratory: Ninety Decisions

1. about: a data commentary

In a data commentary, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check logical direction before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

2. roughly: a comprehension answer

In a comprehension answer, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check reference set before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

3. nearly: a formal email

In a formal email, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check scope before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

4. almost: a research summary

In a research summary, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check logical direction before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

5. up to: a literature response

In a literature response, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check reference set before elegance. Explain the difference in reference set. Do not accept “it sounds better” as the final answer; name the meaning relation.

6. at least: a comparison review

In a comparison review, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check scope before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

7. at most: an argumentative paragraph

In an argumentative paragraph, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check logical direction before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

8. about: a reflective journal

In a reflective journal, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check reference set before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

9. roughly: a history essay

In a history essay, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check scope before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

10. nearly: a speech

In a speech, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check logical direction before elegance. Explain the difference in logical direction. Do not accept “it sounds better” as the final answer; name the meaning relation.

11. almost: a public notice

In a public notice, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check reference set before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

12. up to: a parent-teacher note

In a parent-teacher note, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check scope before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

13. at least: a problem-solution paragraph

In a problem-solution paragraph, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check logical direction before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

14. at most: a school report

In a school report, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check reference set before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

15. about: a narrative scene

In a narrative scene, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check scope before elegance. Explain the difference in scope. Do not accept “it sounds better” as the final answer; name the meaning relation.

16. roughly: a geography response

In a geography response, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check logical direction before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

17. nearly: a project evaluation

In a project evaluation, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check reference set before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

18. almost: a set of instructions

In a set of instructions, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check scope before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

19. up to: an oral answer

In an oral answer, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check logical direction before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

20. at least: a science explanation

In a science explanation, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check reference set before elegance. Explain the difference in reference set. Do not accept “it sounds better” as the final answer; name the meaning relation.

21. at most: a data commentary

In a data commentary, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check scope before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

22. about: a comprehension answer

In a comprehension answer, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check logical direction before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

23. roughly: a formal email

In a formal email, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check reference set before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

24. nearly: a research summary

In a research summary, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check scope before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

25. almost: a literature response

In a literature response, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check logical direction before elegance. Explain the difference in logical direction. Do not accept “it sounds better” as the final answer; name the meaning relation.

26. up to: a comparison review

In a comparison review, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check reference set before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

27. at least: an argumentative paragraph

In an argumentative paragraph, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check scope before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

28. at most: a reflective journal

In a reflective journal, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check logical direction before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

29. about: a history essay

In a history essay, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check reference set before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

30. roughly: a speech

In a speech, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check scope before elegance. Explain the difference in scope. Do not accept “it sounds better” as the final answer; name the meaning relation.

31. nearly: a public notice

In a public notice, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check logical direction before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

32. almost: a parent-teacher note

In a parent-teacher note, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check reference set before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

33. up to: a problem-solution paragraph

In a problem-solution paragraph, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check scope before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

34. at least: a school report

In a school report, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check logical direction before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

35. at most: a narrative scene

In a narrative scene, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check reference set before elegance. Explain the difference in reference set. Do not accept “it sounds better” as the final answer; name the meaning relation.

36. about: a geography response

In a geography response, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check scope before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

37. roughly: a project evaluation

In a project evaluation, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check logical direction before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

38. nearly: a set of instructions

In a set of instructions, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check reference set before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

39. almost: an oral answer

In an oral answer, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check scope before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

40. up to: a science explanation

In a science explanation, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check logical direction before elegance. Explain the difference in logical direction. Do not accept “it sounds better” as the final answer; name the meaning relation.

41. at least: a data commentary

In a data commentary, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check reference set before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

42. at most: a comprehension answer

In a comprehension answer, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check scope before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

43. about: a formal email

In a formal email, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check logical direction before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

44. roughly: a research summary

In a research summary, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check reference set before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

45. nearly: a literature response

In a literature response, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check scope before elegance. Explain the difference in scope. Do not accept “it sounds better” as the final answer; name the meaning relation.

46. almost: a comparison review

In a comparison review, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check logical direction before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

47. up to: an argumentative paragraph

In an argumentative paragraph, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check reference set before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

48. at least: a reflective journal

In a reflective journal, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check scope before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

49. at most: a history essay

In a history essay, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check logical direction before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

50. about: a speech

In a speech, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check reference set before elegance. Explain the difference in reference set. Do not accept “it sounds better” as the final answer; name the meaning relation.

51. roughly: a public notice

In a public notice, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check scope before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

52. nearly: a parent-teacher note

In a parent-teacher note, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check logical direction before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

53. almost: a problem-solution paragraph

In a problem-solution paragraph, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check reference set before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

54. up to: a school report

In a school report, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check scope before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

55. at least: a narrative scene

In a narrative scene, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check logical direction before elegance. Explain the difference in logical direction. Do not accept “it sounds better” as the final answer; name the meaning relation.

56. at most: a geography response

In a geography response, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check reference set before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

57. about: a project evaluation

In a project evaluation, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check scope before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

58. roughly: a set of instructions

In a set of instructions, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check logical direction before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

59. nearly: an oral answer

In an oral answer, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check reference set before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

60. almost: a science explanation

In a science explanation, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check scope before elegance. Explain the difference in scope. Do not accept “it sounds better” as the final answer; name the meaning relation.

61. up to: a data commentary

In a data commentary, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check logical direction before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

62. at least: a comprehension answer

In a comprehension answer, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check reference set before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

63. at most: a formal email

In a formal email, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check scope before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

64. about: a research summary

In a research summary, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check logical direction before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

65. roughly: a literature response

In a literature response, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check reference set before elegance. Explain the difference in reference set. Do not accept “it sounds better” as the final answer; name the meaning relation.

66. nearly: a comparison review

In a comparison review, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check scope before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

67. almost: an argumentative paragraph

In an argumentative paragraph, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check logical direction before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

68. up to: a reflective journal

In a reflective journal, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check reference set before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

69. at least: a history essay

In a history essay, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check scope before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

70. at most: a speech

In a speech, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check logical direction before elegance. Explain the difference in logical direction. Do not accept “it sounds better” as the final answer; name the meaning relation.

71. about: a public notice

In a public notice, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check reference set before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

72. roughly: a parent-teacher note

In a parent-teacher note, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check scope before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

73. nearly: a problem-solution paragraph

In a problem-solution paragraph, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check logical direction before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

74. almost: a school report

In a school report, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check reference set before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

75. up to: a narrative scene

In a narrative scene, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check scope before elegance. Explain the difference in scope. Do not accept “it sounds better” as the final answer; name the meaning relation.

76. at least: a geography response

In a geography response, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check logical direction before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

77. at most: a project evaluation

In a project evaluation, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check reference set before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

78. about: a set of instructions

In a set of instructions, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check scope before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

79. roughly: an oral answer

In an oral answer, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check logical direction before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

80. nearly: a science explanation

In a science explanation, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check reference set before elegance. Explain the difference in reference set. Do not accept “it sounds better” as the final answer; name the meaning relation.

81. almost: a data commentary

In a data commentary, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check scope before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

82. up to: a comprehension answer

In a comprehension answer, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check logical direction before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

83. at least: a formal email

In a formal email, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check reference set before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

84. at most: a research summary

In a research summary, decide whether at most is licensed by the intended meaning. Start from the model “Each answer should contain at most three sentences.” and compare it with “Each answer should contain at most three sentences and must contain four.” and “Each answer should contain at most three sentences.”. Check scope before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

85. about: a literature response

In a literature response, decide whether about is licensed by the intended meaning. Start from the model “About 200 people attended the open house.” and compare it with “Exactly about 200 people attended.” and “About 200 people attended.”. Check logical direction before elegance. Explain the difference in logical direction. Do not accept “it sounds better” as the final answer; name the meaning relation.

86. roughly: a comparison review

In a comparison review, decide whether roughly is licensed by the intended meaning. Start from the model “The first stage took roughly twenty minutes.” and compare it with “The first stage took roughly 20.000 minutes.” and “The first stage took roughly twenty minutes.”. Check reference set before elegance. Paraphrase the relation without the target expression, then put the expression back only if it earns its place.

87. nearly: an argumentative paragraph

In an argumentative paragraph, decide whether nearly is licensed by the intended meaning. Start from the model “Nearly 100 entries were submitted.” and compare it with “Nearly 100 entries were submitted, in fact there were 103.” and “About 100 entries were submitted.”. Check scope before elegance. Write a second example on an unrelated topic. Transfer, not repetition, is the stronger test of control.

88. almost: a reflective journal

In a reflective journal, decide whether almost is licensed by the intended meaning. Start from the model “The tank was almost full when the valve closed.” and compare it with “The tank was almost full and overflowing for an hour.” and “The tank was full and overflowing for an hour.”. Check logical direction before elegance. Ask what a careful reader would infer. If that inference is stronger than the evidence, revise.

89. up to: a history essay

In a history essay, decide whether up to is licensed by the intended meaning. Start from the model “The hall can hold up to 300 people.” and compare it with “The hall holds 300 people at every event because capacity is up to 300.” and “The hall can hold up to 300 people.”. Check reference set before elegance. Compare with the simplest possible wording. Keep the specialist form only when it preserves a useful distinction.

90. at least: a speech

In a speech, decide whether at least is licensed by the intended meaning. Start from the model “Bring at least two examples to the discussion.” and compare it with “Bring at least two examples, but no more than two.” and “Bring exactly two examples.”. Check scope before elegance. Explain the difference in scope. Do not accept “it sounds better” as the final answer; name the meaning relation.

5. Genre Transfer

1. academic essay

In a academic essay, the choice of about should be driven by the relation the sentence needs. It presents a value as an estimate near the stated number. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

2. science report

In a science report, the choice of roughly should be driven by the relation the sentence needs. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

3. history essay

In a history essay, the choice of nearly should be driven by the relation the sentence needs. It indicates that a value or state came close to a target without reaching it in the intended reading. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

4. geography response

In a geography response, the choice of almost should be driven by the relation the sentence needs. It indicates near attainment, completion or occurrence without full attainment. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

5. literature analysis

In a literature analysis, the choice of up to should be driven by the relation the sentence needs. It sets an upper limit on a possible, allowed or observed value. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

6. comprehension answer

In a comprehension answer, the choice of at least should be driven by the relation the sentence needs. It sets a lower bound: the true or required value is no smaller than the stated amount. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

7. creative writing

In a creative writing, the choice of at most should be driven by the relation the sentence needs. It sets an upper bound: the true or permitted value is no greater than the stated amount. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

8. reflective writing

In a reflective writing, the choice of about should be driven by the relation the sentence needs. It presents a value as an estimate near the stated number. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

9. formal email

In a formal email, the choice of roughly should be driven by the relation the sentence needs. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

10. speech

In a speech, the choice of nearly should be driven by the relation the sentence needs. It indicates that a value or state came close to a target without reaching it in the intended reading. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

11. data commentary

In a data commentary, the choice of almost should be driven by the relation the sentence needs. It indicates near attainment, completion or occurrence without full attainment. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

12. project evaluation

In a project evaluation, the choice of up to should be driven by the relation the sentence needs. It sets an upper limit on a possible, allowed or observed value. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

13. research summary

In a research summary, the choice of at least should be driven by the relation the sentence needs. It sets a lower bound: the true or required value is no smaller than the stated amount. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

14. instructions

In a instructions, the choice of at most should be driven by the relation the sentence needs. It sets an upper bound: the true or permitted value is no greater than the stated amount. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

15. public notice

In a public notice, the choice of about should be driven by the relation the sentence needs. It presents a value as an estimate near the stated number. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

16. oral presentation

In a oral presentation, the choice of roughly should be driven by the relation the sentence needs. It marks an approximate estimate and often foregrounds that the figure is coarse rather than exact. Write one accurate version, one deliberate near miss and one plain-English paraphrase. The three versions force the learner to separate the underlying function from the surface vocabulary.

6. Error Clinic

1. Diagnose the intended relation first

When nearly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test register, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

2. Diagnose the intended relation first

When up to looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test punctuation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

3. Diagnose the intended relation first

When at most looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test claim strength, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

4. Diagnose the intended relation first

When roughly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test reader expectation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

5. Diagnose the intended relation first

When almost looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test logical direction, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

6. Diagnose the intended relation first

When at least looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test ambiguity, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

7. Diagnose the intended relation first

When about looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test evidence, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

8. Diagnose the intended relation first

When nearly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test sentence rhythm, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

9. Diagnose the intended relation first

When up to looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test cohesion, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

10. Diagnose the intended relation first

When at most looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test reference set, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

11. Diagnose the intended relation first

When roughly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test threshold, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

12. Diagnose the intended relation first

When almost looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test frequency interpretation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

13. Diagnose the intended relation first

When at least looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test precision, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

14. Diagnose the intended relation first

When about looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test tone, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

15. Diagnose the intended relation first

When nearly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test scope, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

16. Diagnose the intended relation first

When up to looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test register, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

17. Diagnose the intended relation first

When at most looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test punctuation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

18. Diagnose the intended relation first

When roughly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test claim strength, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

19. Diagnose the intended relation first

When almost looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test reader expectation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

20. Diagnose the intended relation first

When at least looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test logical direction, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

21. Diagnose the intended relation first

When about looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test ambiguity, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

22. Diagnose the intended relation first

When nearly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test evidence, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

23. Diagnose the intended relation first

When up to looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test sentence rhythm, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

24. Diagnose the intended relation first

When at most looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test cohesion, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

25. Diagnose the intended relation first

When roughly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test reference set, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

26. Diagnose the intended relation first

When almost looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test threshold, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

27. Diagnose the intended relation first

When at least looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test frequency interpretation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

28. Diagnose the intended relation first

When about looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test precision, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

29. Diagnose the intended relation first

When nearly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test tone, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

30. Diagnose the intended relation first

When up to looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test scope, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

31. Diagnose the intended relation first

When at most looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test register, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

32. Diagnose the intended relation first

When roughly looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test punctuation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

33. Diagnose the intended relation first

When almost looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test claim strength, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

34. Diagnose the intended relation first

When at least looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test reader expectation, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

35. Diagnose the intended relation first

When about looks wrong, do not begin by swapping vocabulary. State what the writer means, identify the grammatical structure, test logical direction, compare the nearest alternative and then revise. A locally fluent replacement can still damage the paragraph if it changes the logic. Correction should restore the intended relation, not merely produce a sentence that sounds more formal.

7. Retrieval Workbook

1. Build, replace, explain

Write a sentence in a speech that genuinely needs almost. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

2. Build, replace, explain

Write a sentence in a project evaluation that genuinely needs up to. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

3. Build, replace, explain

Write a sentence in a research summary that genuinely needs at least. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

4. Build, replace, explain

Write a sentence in a public notice that genuinely needs at most. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

5. Build, replace, explain

Write a sentence in a set of instructions that genuinely needs about. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

6. Build, replace, explain

Write a sentence in a literature response that genuinely needs roughly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

7. Build, replace, explain

Write a sentence in a parent-teacher note that genuinely needs nearly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

8. Build, replace, explain

Write a sentence in an oral answer that genuinely needs almost. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

9. Build, replace, explain

Write a sentence in a comparison review that genuinely needs up to. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

10. Build, replace, explain

Write a sentence in a problem-solution paragraph that genuinely needs at least. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

11. Build, replace, explain

Write a sentence in a science explanation that genuinely needs at most. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

12. Build, replace, explain

Write a sentence in an argumentative paragraph that genuinely needs about. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

13. Build, replace, explain

Write a sentence in a school report that genuinely needs roughly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

14. Build, replace, explain

Write a sentence in a data commentary that genuinely needs nearly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

15. Build, replace, explain

Write a sentence in a reflective journal that genuinely needs almost. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

16. Build, replace, explain

Write a sentence in a narrative scene that genuinely needs up to. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

17. Build, replace, explain

Write a sentence in a comprehension answer that genuinely needs at least. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

18. Build, replace, explain

Write a sentence in a history essay that genuinely needs at most. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

19. Build, replace, explain

Write a sentence in a geography response that genuinely needs about. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

20. Build, replace, explain

Write a sentence in a formal email that genuinely needs roughly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

21. Build, replace, explain

Write a sentence in a speech that genuinely needs nearly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

22. Build, replace, explain

Write a sentence in a project evaluation that genuinely needs almost. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

23. Build, replace, explain

Write a sentence in a research summary that genuinely needs up to. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

24. Build, replace, explain

Write a sentence in a public notice that genuinely needs at least. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

25. Build, replace, explain

Write a sentence in a set of instructions that genuinely needs at most. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

26. Build, replace, explain

Write a sentence in a literature response that genuinely needs about. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

27. Build, replace, explain

Write a sentence in a parent-teacher note that genuinely needs roughly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

28. Build, replace, explain

Write a sentence in an oral answer that genuinely needs nearly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

29. Build, replace, explain

Write a sentence in a comparison review that genuinely needs almost. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

30. Build, replace, explain

Write a sentence in a problem-solution paragraph that genuinely needs up to. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

31. Build, replace, explain

Write a sentence in a science explanation that genuinely needs at least. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

32. Build, replace, explain

Write a sentence in an argumentative paragraph that genuinely needs at most. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

33. Build, replace, explain

Write a sentence in a school report that genuinely needs about. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

34. Build, replace, explain

Write a sentence in a data commentary that genuinely needs roughly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

35. Build, replace, explain

Write a sentence in a reflective journal that genuinely needs nearly. Replace it with the nearest alternative. Under the two sentences, write four short labels: intended relation, grammar, reader inference and register. Finish with a third example on a different subject. The exercise is complete only when the explanation predicts why the alternatives behave differently.

8. Teaching and Independent Practice

For independent study, collect real sentences rather than isolated words. Mark the decision point, predict the function, check the surrounding clause or phrase, and record one contrastive example. Revisit it later by rebuilding rather than rereading. The point is to retrieve the distinction, not merely recognise the page.

For teaching, ask replacement questions: “What changes if we use the other form?” “Does the sentence become stronger, weaker, broader, narrower or simply different?” “What grammar has to change?” These questions expose partial knowledge quickly and make feedback more useful than a correct/incorrect mark alone.

9. Further Study

For general reference, see the British Council English grammar reference and Cambridge Dictionary grammar reference. Approximation connects directly to Frequency and Habit when vague frequency needs to become a bounded or estimated quantity, to Proportion and Ratio when the underlying relation is numerical, to Comparison and Degree when the issue is a scale or threshold, and to Change and Trend when estimates describe movement over time. The series hub provides the wider precision map.

10. Final Checklist

1. What relation am I expressing?

Final check: 1. What relation am I expressing? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

2. What does the form grammatically attach to?

Final check: 2. What does the form grammatically attach to? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

3. What scope does it have?

Final check: 3. What scope does it have? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

4. What would the nearest alternative change?

Final check: 4. What would the nearest alternative change? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

5. Is the register appropriate?

Final check: 5. Is the register appropriate? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

6. Does the evidence justify the wording?

Final check: 6. Does the evidence justify the wording? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

7. Could the reader infer more than I intend?

Final check: 7. Could the reader infer more than I intend? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

8. Can I paraphrase the relation plainly?

Final check: 8. Can I paraphrase the relation plainly? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

9. Can I use the same function in a new context?

Final check: 9. Can I use the same function in a new context? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

10. Is the simplest accurate form already enough?

Final check: 10. Is the simplest accurate form already enough? State the intended relationship, identify the nearest competing expression, and explain one concrete reason the chosen wording is more accurate. If the answer depends only on “it sounds better”, revise again.

Advanced Application: Transfer Problems

Application 1: about — a science explanation

In a science explanation, write one sentence that genuinely needs about. Then replace it with the closest alternative and explain what changes in estimate tolerance. Move the same distinction into a comprehension answer without copying the original wording. Finish by naming one context in which about would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 2: roughly — an oral response

Build a three-sentence mini-example in an oral response: first establish the context, then use roughly, then write a correction that would be needed if upper versus lower bound changed. Rework the example for a geography explanation. This tests whether roughly is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 3: nearly — a public notice

Imagine a careful editor questions nearly in a public notice. Defend it by identifying the intended relation and the evidence that supports near attainment. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a classroom discussion. A strong answer makes the boundary of use visible as well as the successful use.

Application 4: almost — a reflective paragraph

Write two versions of the same idea in a reflective paragraph: one with almost, one without it. Compare the reader’s likely interpretation, paying particular attention to false precision. Then change the audience and rewrite the idea for a report conclusion. The goal is not to preserve the word; it is to preserve the meaning. Keep almost only when it earns its place.

Application 5: up to — a short speech

Create a deliberate near miss with up to in a short speech, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention maximum versus typical value. Then create a fresh correct example in a history paragraph. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 6: at least — a data commentary

In a data commentary, write one sentence that genuinely needs at least. Then replace it with the closest alternative and explain what changes in rounding. Move the same distinction into a project evaluation without copying the original wording. Finish by naming one context in which at least would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 7: at most — a narrative scene

Build a three-sentence mini-example in a narrative scene: first establish the context, then use at most, then write a correction that would be needed if whether the target has actually been reached changed. Rework the example for a research summary. This tests whether at most is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 8: about — a literature response

Imagine a careful editor questions about in a literature response. Defend it by identifying the intended relation and the evidence that supports estimate tolerance. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a technical note. A strong answer makes the boundary of use visible as well as the successful use.

Application 9: roughly — a comparison review

Write two versions of the same idea in a comparison review: one with roughly, one without it. Compare the reader’s likely interpretation, paying particular attention to upper versus lower bound. Then change the audience and rewrite the idea for a problem-solution paragraph. The goal is not to preserve the word; it is to preserve the meaning. Keep roughly only when it earns its place.

Application 10: nearly — a revision note

Create a deliberate near miss with nearly in a revision note, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention near attainment. Then create a fresh correct example in a policy note. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 11: almost — a formal email

In a formal email, write one sentence that genuinely needs almost. Then replace it with the closest alternative and explain what changes in false precision. Move the same distinction into a parent update without copying the original wording. Finish by naming one context in which almost would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 12: up to — a comprehension answer

Build a three-sentence mini-example in a comprehension answer: first establish the context, then use up to, then write a correction that would be needed if maximum versus typical value changed. Rework the example for a set of instructions. This tests whether up to is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 13: at least — a geography explanation

Imagine a careful editor questions at least in a geography explanation. Defend it by identifying the intended relation and the evidence that supports rounding. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in an exam response. A strong answer makes the boundary of use visible as well as the successful use.

Application 14: at most — a classroom discussion

Write two versions of the same idea in a classroom discussion: one with at most, one without it. Compare the reader’s likely interpretation, paying particular attention to whether the target has actually been reached. Then change the audience and rewrite the idea for a science explanation. The goal is not to preserve the word; it is to preserve the meaning. Keep at most only when it earns its place.

Application 15: about — a report conclusion

Create a deliberate near miss with about in a report conclusion, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention estimate tolerance. Then create a fresh correct example in an oral response. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 16: roughly — a history paragraph

In a history paragraph, write one sentence that genuinely needs roughly. Then replace it with the closest alternative and explain what changes in upper versus lower bound. Move the same distinction into a public notice without copying the original wording. Finish by naming one context in which roughly would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 17: nearly — a project evaluation

Build a three-sentence mini-example in a project evaluation: first establish the context, then use nearly, then write a correction that would be needed if near attainment changed. Rework the example for a reflective paragraph. This tests whether nearly is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 18: almost — a research summary

Imagine a careful editor questions almost in a research summary. Defend it by identifying the intended relation and the evidence that supports false precision. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a short speech. A strong answer makes the boundary of use visible as well as the successful use.

Application 19: up to — a technical note

Write two versions of the same idea in a technical note: one with up to, one without it. Compare the reader’s likely interpretation, paying particular attention to maximum versus typical value. Then change the audience and rewrite the idea for a data commentary. The goal is not to preserve the word; it is to preserve the meaning. Keep up to only when it earns its place.

Application 20: at least — a problem-solution paragraph

Create a deliberate near miss with at least in a problem-solution paragraph, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention rounding. Then create a fresh correct example in a narrative scene. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 21: at most — a policy note

In a policy note, write one sentence that genuinely needs at most. Then replace it with the closest alternative and explain what changes in whether the target has actually been reached. Move the same distinction into a literature response without copying the original wording. Finish by naming one context in which at most would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 22: about — a parent update

Build a three-sentence mini-example in a parent update: first establish the context, then use about, then write a correction that would be needed if estimate tolerance changed. Rework the example for a comparison review. This tests whether about is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 23: roughly — a set of instructions

Imagine a careful editor questions roughly in a set of instructions. Defend it by identifying the intended relation and the evidence that supports upper versus lower bound. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a revision note. A strong answer makes the boundary of use visible as well as the successful use.

Application 24: nearly — an exam response

Write two versions of the same idea in an exam response: one with nearly, one without it. Compare the reader’s likely interpretation, paying particular attention to near attainment. Then change the audience and rewrite the idea for a formal email. The goal is not to preserve the word; it is to preserve the meaning. Keep nearly only when it earns its place.

Application 25: almost — a science explanation

Create a deliberate near miss with almost in a science explanation, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention false precision. Then create a fresh correct example in a comprehension answer. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 26: up to — an oral response

In an oral response, write one sentence that genuinely needs up to. Then replace it with the closest alternative and explain what changes in maximum versus typical value. Move the same distinction into a geography explanation without copying the original wording. Finish by naming one context in which up to would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 27: at least — a public notice

Build a three-sentence mini-example in a public notice: first establish the context, then use at least, then write a correction that would be needed if rounding changed. Rework the example for a classroom discussion. This tests whether at least is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 28: at most — a reflective paragraph

Imagine a careful editor questions at most in a reflective paragraph. Defend it by identifying the intended relation and the evidence that supports whether the target has actually been reached. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a report conclusion. A strong answer makes the boundary of use visible as well as the successful use.

Application 29: about — a short speech

Write two versions of the same idea in a short speech: one with about, one without it. Compare the reader’s likely interpretation, paying particular attention to estimate tolerance. Then change the audience and rewrite the idea for a history paragraph. The goal is not to preserve the word; it is to preserve the meaning. Keep about only when it earns its place.

Application 30: roughly — a data commentary

Create a deliberate near miss with roughly in a data commentary, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention upper versus lower bound. Then create a fresh correct example in a project evaluation. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 31: nearly — a narrative scene

In a narrative scene, write one sentence that genuinely needs nearly. Then replace it with the closest alternative and explain what changes in near attainment. Move the same distinction into a research summary without copying the original wording. Finish by naming one context in which nearly would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 32: almost — a literature response

Build a three-sentence mini-example in a literature response: first establish the context, then use almost, then write a correction that would be needed if false precision changed. Rework the example for a technical note. This tests whether almost is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 33: up to — a comparison review

Imagine a careful editor questions up to in a comparison review. Defend it by identifying the intended relation and the evidence that supports maximum versus typical value. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a problem-solution paragraph. A strong answer makes the boundary of use visible as well as the successful use.

Application 34: at least — a revision note

Write two versions of the same idea in a revision note: one with at least, one without it. Compare the reader’s likely interpretation, paying particular attention to rounding. Then change the audience and rewrite the idea for a policy note. The goal is not to preserve the word; it is to preserve the meaning. Keep at least only when it earns its place.

Application 35: at most — a formal email

Create a deliberate near miss with at most in a formal email, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention whether the target has actually been reached. Then create a fresh correct example in a parent update. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 36: about — a comprehension answer

In a comprehension answer, write one sentence that genuinely needs about. Then replace it with the closest alternative and explain what changes in estimate tolerance. Move the same distinction into a set of instructions without copying the original wording. Finish by naming one context in which about would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 37: roughly — a geography explanation

Build a three-sentence mini-example in a geography explanation: first establish the context, then use roughly, then write a correction that would be needed if upper versus lower bound changed. Rework the example for an exam response. This tests whether roughly is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 38: nearly — a classroom discussion

Imagine a careful editor questions nearly in a classroom discussion. Defend it by identifying the intended relation and the evidence that supports near attainment. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a science explanation. A strong answer makes the boundary of use visible as well as the successful use.

Application 39: almost — a report conclusion

Write two versions of the same idea in a report conclusion: one with almost, one without it. Compare the reader’s likely interpretation, paying particular attention to false precision. Then change the audience and rewrite the idea for an oral response. The goal is not to preserve the word; it is to preserve the meaning. Keep almost only when it earns its place.

Application 40: up to — a history paragraph

Create a deliberate near miss with up to in a history paragraph, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention maximum versus typical value. Then create a fresh correct example in a public notice. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 41: at least — a project evaluation

In a project evaluation, write one sentence that genuinely needs at least. Then replace it with the closest alternative and explain what changes in rounding. Move the same distinction into a reflective paragraph without copying the original wording. Finish by naming one context in which at least would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 42: at most — a research summary

Build a three-sentence mini-example in a research summary: first establish the context, then use at most, then write a correction that would be needed if whether the target has actually been reached changed. Rework the example for a short speech. This tests whether at most is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 43: about — a technical note

Imagine a careful editor questions about in a technical note. Defend it by identifying the intended relation and the evidence that supports estimate tolerance. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a data commentary. A strong answer makes the boundary of use visible as well as the successful use.

Application 44: roughly — a problem-solution paragraph

Write two versions of the same idea in a problem-solution paragraph: one with roughly, one without it. Compare the reader’s likely interpretation, paying particular attention to upper versus lower bound. Then change the audience and rewrite the idea for a narrative scene. The goal is not to preserve the word; it is to preserve the meaning. Keep roughly only when it earns its place.

Application 45: nearly — a policy note

Create a deliberate near miss with nearly in a policy note, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention near attainment. Then create a fresh correct example in a literature response. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 46: almost — a parent update

In a parent update, write one sentence that genuinely needs almost. Then replace it with the closest alternative and explain what changes in false precision. Move the same distinction into a comparison review without copying the original wording. Finish by naming one context in which almost would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 47: up to — a set of instructions

Build a three-sentence mini-example in a set of instructions: first establish the context, then use up to, then write a correction that would be needed if maximum versus typical value changed. Rework the example for a revision note. This tests whether up to is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 48: at least — an exam response

Imagine a careful editor questions at least in an exam response. Defend it by identifying the intended relation and the evidence that supports rounding. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a formal email. A strong answer makes the boundary of use visible as well as the successful use.

Application 49: at most — a science explanation

Write two versions of the same idea in a science explanation: one with at most, one without it. Compare the reader’s likely interpretation, paying particular attention to whether the target has actually been reached. Then change the audience and rewrite the idea for a comprehension answer. The goal is not to preserve the word; it is to preserve the meaning. Keep at most only when it earns its place.

Application 50: about — an oral response

Create a deliberate near miss with about in an oral response, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention estimate tolerance. Then create a fresh correct example in a geography explanation. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 51: roughly — a public notice

In a public notice, write one sentence that genuinely needs roughly. Then replace it with the closest alternative and explain what changes in upper versus lower bound. Move the same distinction into a classroom discussion without copying the original wording. Finish by naming one context in which roughly would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 52: nearly — a reflective paragraph

Build a three-sentence mini-example in a reflective paragraph: first establish the context, then use nearly, then write a correction that would be needed if near attainment changed. Rework the example for a report conclusion. This tests whether nearly is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 53: almost — a short speech

Imagine a careful editor questions almost in a short speech. Defend it by identifying the intended relation and the evidence that supports false precision. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a history paragraph. A strong answer makes the boundary of use visible as well as the successful use.

Application 54: up to — a data commentary

Write two versions of the same idea in a data commentary: one with up to, one without it. Compare the reader’s likely interpretation, paying particular attention to maximum versus typical value. Then change the audience and rewrite the idea for a project evaluation. The goal is not to preserve the word; it is to preserve the meaning. Keep up to only when it earns its place.

Application 55: at least — a narrative scene

Create a deliberate near miss with at least in a narrative scene, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention rounding. Then create a fresh correct example in a research summary. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 56: at most — a literature response

In a literature response, write one sentence that genuinely needs at most. Then replace it with the closest alternative and explain what changes in whether the target has actually been reached. Move the same distinction into a technical note without copying the original wording. Finish by naming one context in which at most would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 57: about — a comparison review

Build a three-sentence mini-example in a comparison review: first establish the context, then use about, then write a correction that would be needed if estimate tolerance changed. Rework the example for a problem-solution paragraph. This tests whether about is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 58: roughly — a revision note

Imagine a careful editor questions roughly in a revision note. Defend it by identifying the intended relation and the evidence that supports upper versus lower bound. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a policy note. A strong answer makes the boundary of use visible as well as the successful use.

Application 59: nearly — a formal email

Write two versions of the same idea in a formal email: one with nearly, one without it. Compare the reader’s likely interpretation, paying particular attention to near attainment. Then change the audience and rewrite the idea for a parent update. The goal is not to preserve the word; it is to preserve the meaning. Keep nearly only when it earns its place.

Application 60: almost — a comprehension answer

Create a deliberate near miss with almost in a comprehension answer, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention false precision. Then create a fresh correct example in a set of instructions. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 61: up to — a geography explanation

In a geography explanation, write one sentence that genuinely needs up to. Then replace it with the closest alternative and explain what changes in maximum versus typical value. Move the same distinction into an exam response without copying the original wording. Finish by naming one context in which up to would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 62: at least — a classroom discussion

Build a three-sentence mini-example in a classroom discussion: first establish the context, then use at least, then write a correction that would be needed if rounding changed. Rework the example for a science explanation. This tests whether at least is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 63: at most — a report conclusion

Imagine a careful editor questions at most in a report conclusion. Defend it by identifying the intended relation and the evidence that supports whether the target has actually been reached. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in an oral response. A strong answer makes the boundary of use visible as well as the successful use.

Application 64: about — a history paragraph

Write two versions of the same idea in a history paragraph: one with about, one without it. Compare the reader’s likely interpretation, paying particular attention to estimate tolerance. Then change the audience and rewrite the idea for a public notice. The goal is not to preserve the word; it is to preserve the meaning. Keep about only when it earns its place.

Application 65: roughly — a project evaluation

Create a deliberate near miss with roughly in a project evaluation, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention upper versus lower bound. Then create a fresh correct example in a reflective paragraph. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 66: nearly — a research summary

In a research summary, write one sentence that genuinely needs nearly. Then replace it with the closest alternative and explain what changes in near attainment. Move the same distinction into a short speech without copying the original wording. Finish by naming one context in which nearly would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 67: almost — a technical note

Build a three-sentence mini-example in a technical note: first establish the context, then use almost, then write a correction that would be needed if false precision changed. Rework the example for a data commentary. This tests whether almost is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 68: up to — a problem-solution paragraph

Imagine a careful editor questions up to in a problem-solution paragraph. Defend it by identifying the intended relation and the evidence that supports maximum versus typical value. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a narrative scene. A strong answer makes the boundary of use visible as well as the successful use.

Application 69: at least — a policy note

Write two versions of the same idea in a policy note: one with at least, one without it. Compare the reader’s likely interpretation, paying particular attention to rounding. Then change the audience and rewrite the idea for a literature response. The goal is not to preserve the word; it is to preserve the meaning. Keep at least only when it earns its place.

Application 70: at most — a parent update

Create a deliberate near miss with at most in a parent update, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention whether the target has actually been reached. Then create a fresh correct example in a comparison review. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 71: about — a set of instructions

In a set of instructions, write one sentence that genuinely needs about. Then replace it with the closest alternative and explain what changes in estimate tolerance. Move the same distinction into a revision note without copying the original wording. Finish by naming one context in which about would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 72: roughly — an exam response

Build a three-sentence mini-example in an exam response: first establish the context, then use roughly, then write a correction that would be needed if upper versus lower bound changed. Rework the example for a formal email. This tests whether roughly is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 73: nearly — a science explanation

Imagine a careful editor questions nearly in a science explanation. Defend it by identifying the intended relation and the evidence that supports near attainment. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a comprehension answer. A strong answer makes the boundary of use visible as well as the successful use.

Application 74: almost — an oral response

Write two versions of the same idea in an oral response: one with almost, one without it. Compare the reader’s likely interpretation, paying particular attention to false precision. Then change the audience and rewrite the idea for a geography explanation. The goal is not to preserve the word; it is to preserve the meaning. Keep almost only when it earns its place.

Application 75: up to — a public notice

Create a deliberate near miss with up to in a public notice, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention maximum versus typical value. Then create a fresh correct example in a classroom discussion. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Application 76: at least — a reflective paragraph

In a reflective paragraph, write one sentence that genuinely needs at least. Then replace it with the closest alternative and explain what changes in rounding. Move the same distinction into a report conclusion without copying the original wording. Finish by naming one context in which at least would mislead the reader. The task is successful only if the explanation states the semantic difference rather than appealing to style.

Application 77: at most — a short speech

Build a three-sentence mini-example in a short speech: first establish the context, then use at most, then write a correction that would be needed if whether the target has actually been reached changed. Rework the example for a history paragraph. This tests whether at most is tied to a memorised sentence or to a transferable decision about approximation and precision.

Application 78: about — a data commentary

Imagine a careful editor questions about in a data commentary. Defend it by identifying the intended relation and the evidence that supports estimate tolerance. Next, concede a case where the editor would be right and supply a better alternative. Repeat the decision in a project evaluation. A strong answer makes the boundary of use visible as well as the successful use.

Application 79: roughly — a narrative scene

Write two versions of the same idea in a narrative scene: one with roughly, one without it. Compare the reader’s likely interpretation, paying particular attention to upper versus lower bound. Then change the audience and rewrite the idea for a research summary. The goal is not to preserve the word; it is to preserve the meaning. Keep roughly only when it earns its place.

Application 80: nearly — a literature response

Create a deliberate near miss with nearly in a literature response, diagnose exactly why it fails, and repair it with the smallest possible change. Your diagnosis must mention near attainment. Then create a fresh correct example in a technical note. This repair-first method turns approximation and precision into an editing skill rather than a recognition exercise.

Closing Principle

Approximation is precise when it represents the actual resolution of knowledge. Bounds are precise when they tell the reader what cannot be lower or higher. Do not manufacture certainty with extra digits or hide a maximum behind an average-sounding phrase.