Foreign Secretary, resigned days after the Argentine invasion of the Falklands even though he was not personally to blame - the classic case of a minister taking responsibility for departmental failure. Thatcher tried to keep him; he insisted on going.
Still cited in every essay because modern equivalents are missing - which is itself the point.
Carrington had not misled anyone, broken any rule or behaved badly - there was no Ministerial Code to breach in 1982 and no personal allegation against him. That absence is what makes the case the clean benchmark for the departmental half of the doctrine.
The Foreign Office had misread Argentine intentions before the invasion of the Falklands in April 1982, and Carrington resigned within days for his department's failure - not his own. No Foreign or Home Secretary has resigned on those pure departmental grounds since, which is why every essay still reaches back to 1982.
There was no cabinet split and no public dissent - Carrington supported the government line to the end. Keep this case out of CMR paragraphs; its work is all on the individual side.
The Ministerial Code was first published in 1992, a decade after Carrington went. His resignation was enforced by something older - a sense of honour the modern enforcement machinery was later built to replace, and arguably never has.
Thatcher tried to persuade him to stay; he insisted on going. Contrast Patel 2020, where the PM's protection decided the opposite outcome - the pair of cases shows the convention shifting from minister's honour to Prime Minister's calculation.
Nobody forced the resignation - no inquiry, no adviser, no vote. The convention worked unprompted, which is exactly what no modern departmental case (Williamson 2020, May's Home Office record 2010-16) has repeated.
Use Carrington as the benchmark, then measure the modern cases against it: the standard exam move is Carrington against Williamson, 1982 against 2020, honour against survival.
Leader of the Commons, resigned because he could not publicly support the Iraq War - the textbook constructive resignation under collective responsibility, complete with a resignation speech to the Commons that essays can quote.
Cook's conduct was never in question - no Code breach, no scandal, no inquiry. His resignation was a matter of conscience, which is what separates constructive resignation from every IMR case on this grid.
Cook had left the Foreign Office in 2001 and resigned from the post of Leader of the Commons - no department, no departmental failure. The case does no work on the IMR side.
Cook could not publicly support the invasion of Iraq, so in March 2003 he left the cabinet and said why - the doctrine operating exactly as written: support the line in public or go. His resignation speech to the Commons received an unprecedented standing ovation and remains the textbook constructive resignation.
Nothing in the Ministerial Code was engaged - the Code polices conduct, not conscience. Clare Short's delayed exit two months later (citing the absence of UN authority for reconstruction) makes the instructive contrast: the doctrine allows a clean exit or an awkward one.
Blair could not stop him and did not try to sack him - the choice belonged entirely to the minister. CMR's resignation half is minister-enforced; its sacking half (Braverman, November 2023) is PM-enforced.
The convention bit exactly as designed: a minister who could not support the line left office rather than break unity in post. Pair with July 2022, where the same doctrine bit in the opposite direction - against the Prime Minister.
Quote the resignation speech's logic - he could not support a war 'without international agreement or domestic support' - and you have AO1, AO2 and a dated example in one sentence.
Home Secretary, resigned over Windrush after inadvertently misleading the Commons about removal targets. The underlying failure was departmental, but the resignation trigger was personal: misleading Parliament is the Code's clearest red line.
The standard modern proof that IMR-personal still bites.
Rudd told the Home Affairs Select Committee there were no removal targets; documents then showed targets existed. She resigned in April 2018 for inadvertently misleading the Commons - the personal trigger, not the underlying Windrush scandal itself.
The Windrush scandal was a departmental disaster years in the making - but it was the misleading of Parliament, not the scandal, that forced the resignation. The case proves the modern rule: departments fail and ministers survive; misleading the Commons still kills.
No collective element - Rudd defended government policy throughout. The case lives entirely on the individual side of the doctrine.
Misleading Parliament is the Code's clearest red line - the one breach where resignation remains close to automatic, because Parliament's ability to hold ministers to account depends on being told the truth.
May defended her for days - but the select committee evidence trail made the position unrecoverable. Where evidence is public and parliamentary, prime ministerial protection has limits; where it stays internal (Patel 2020), protection works.
From the committee evidence emerging to resignation took days, not months. When the trigger is personal and parliamentary, the convention still moves fast.
The standard modern proof that IMR-personal still bites - and the half of the Windrush story that pairs with May's own survival of the Home Office years for the non-resignation column.
Home Secretary. The independent inquiry upheld bullying complaints - a Code breach. Johnson kept her in post, and the Independent Adviser on Ministers' Interests, Sir Alex Allan, resigned instead.
The case that shows the Code is only as strong as the Prime Minister's will to enforce it.
Sir Alex Allan's independent inquiry found Patel's conduct towards officials had breached the Code, even if unintentionally - bullying is personal conduct, squarely the Code's territory. The finding was clear; the consequence never came.
No departmental failure was alleged - the Home Office's troubles in this period belong to other cases. Keep Patel for the enforcement argument, not the departmental one.
No collective element - Patel never broke the government line. (Her separate 2017 resignation under May, for unauthorised meetings with Israeli officials, is a distinct IMR-personal case - and do not confuse her with Braverman, a different Home Secretary, whose 2022 security-breach resignation under Truss is its own row in the examples library.)
The machinery worked perfectly up to the final step: independent adviser investigates, finds a breach, reports - and the Prime Minister, sole judge under the Code, overruled him. Sir Alex Allan resigned instead of the minister; his successor Lord Geidt resigned in 2022 over the same structural problem.
Johnson's protection was the entire outcome: same finding, different PM, different result. The Code has exactly the force the Prime Minister of the day gives it - this case is that sentence in evidence form.
A found breach with no consequence for the minister - the clearest modern case of the convention not biting. The adviser's resignation in the minister's place is the detail examiners reward, because it shows where the system's pressure actually escaped.
The case that turns a descriptive essay analytical: enforcement is the PM's choice, so IMR's bite varies with prime ministerial interest - which links this grid to PM-power questions.
The PM power grid's Party Control and Big Beasts columns are the other half of that argument: the same calculation that protects a minister can also remove a Prime Minister.
Education Secretary during the A-level grading algorithm fiasco - a major departmental failure on his watch, with no resignation. The standout modern case of IMR-departmental failing to enforce.
Pair with Carrington 1982 for the died-or-evolved debate in a single paragraph.
No personal misconduct was alleged in 2020 - no misleading of Parliament, no Code breach. (His November 2022 resignation under Sunak, over bullying allegations, is a separate personal-conduct episode - keep the two apart.)
The 2020 A-level algorithm fiasco - grades downgraded, public outcry, a forced U-turn - happened squarely on the Education Secretary's watch. On the Carrington standard he would have gone; he stayed for over a year and was removed in the September 2021 reshuffle, not by the convention.
No collective element - the government line held throughout. The case does all its work on the departmental side of IMR.
Departmental failure is not a Code breach - the Code polices personal conduct, which is precisely the gap this case exposes: the modern machinery has no mechanism at all for the Carrington scenario.
Johnson kept him through the storm. Like Patel in the same year, the outcome was decided in Downing Street, not by the convention - two cases, one PM, one pattern.
The standout modern case of IMR-departmental not biting - put it next to May surviving repeated Home Office failures 2010-16 and the pattern is established, not anecdotal.
Pair with Carrington for the died-or-evolved debate in a single paragraph: if you argue the convention is dead, explain Rudd; if you argue it lives, explain Williamson.
On 5 July 2022 Rishi Sunak and Sajid Javid resigned from Boris Johnson's Cabinet within minutes of each other over his handling of the Pincher affair. More than fifty ministers and aides followed within 48 hours; Johnson resigned within two days.
This is collective responsibility at full force - Cabinet ministers withdrawing support and ending a premiership.
Neither resigning minister was accused of anything - their own conduct was never the issue. The trigger was the Prime Minister's handling of the Pincher affair, which makes this a pure collective case.
No departmental failure - the Treasury and Health were not the story. The resignations were aimed at the Prime Minister, not at their own records.
On 5 July 2022 Sunak and Javid resigned within minutes of each other, withdrawing collective support and saying so publicly. More than fifty ministers and aides followed within 48 hours - the doctrine's one-voice logic running in reverse: when the voices stop, the government stops.
No Code machinery was engaged - no inquiry, no adviser. The episode shows the conventions' political core: what removed Johnson was the withdrawal of consent, not a rulebook.
The PM was the object, not the decider: Johnson tried to govern on, appointed replacements, and resigned within two days anyway. Contrast every other row, where the PM's choice decided the minister's fate - here the ministers decided the PM's.
CMR at full force: mass resignation removed a sitting Prime Minister in 48 hours. The same convention that looked dead in the leak-ridden May and Truss years proved it can still end a premiership - the strongest single disagree point in the no-longer-operates essay.
Use it as the hinge of any CMR paragraph: suspensions (1975, 2016) and routine leaking argue decay; July 2022 argues the doctrine's teeth are intact and merely conditional.
Resigned as Deputy Prime Minister and Labour deputy leader in autumn 2025 over a tax dispute - a serving Deputy PM resigning is rare and high-profile. The disagree side reads the case as the doctrine working: serious breach met serious sanction.
The 2025 examiner report rewarded contemporary evidence - this is the most recent named case on the grid.
Rayner resigned as Deputy Prime Minister and Labour deputy leader in autumn 2025 over a tax dispute - personal conduct under the Code, at the highest level of government. A serving Deputy PM resigning is rare enough to be an automatic exam example.
No departmental failure was involved - the issue was personal tax affairs, not the work of her departments. Keep the case on the personal-conduct side.
Not a unity case - though the routine briefing and leaking that continued under Starmer belongs in the cabinet-leaks paragraph of a CMR essay, as evidence that discussions leak under every modern PM.
The Code's machinery produced a resignation at the top of government - the disagree side reads it exactly as the predicted pack does: serious breach met serious sanction, proof the doctrine still operates when the facts are clear.
The breach finding left little room for protection - unlike Patel 2020, where the finding was arguably stronger and the protection held anyway. The comparison between the two cases is a ready-made AO2 paragraph on what actually decides outcomes.
A serving Deputy PM resigning is the convention biting at the highest level it can reach short of the PM - and within the first eighteen months of a government with a landslide majority.
The 2025 examiner report rewarded contemporary evidence over general theory, and this is the most recent named case on the grid - end a paragraph with it.