United States
—
— → — · —
Map A · who and where
Geographic atlas
Western Europe, enlarged
Map B · how similar
UN Security Council · the five permanent members · 1991–2024
To what extent have the five permanent members of the UN Security Council shown commitment to the multilateral order in responding to violent inter-state conflicts, as seen in their rhetoric and behaviour at the Council from 1991 to 2024, and how has that commitment changed over time?
Six methods of analysis, one question. Pick a method on the board to open it.
UN Security Council · six methods, one question · 1991–2024
Research questionTo what extent have the five permanent members shown commitment to the multilateral order in responding to violent inter-state conflicts, and how has that commitment changed?
None of the five has walked away from the UN order: they keep meeting, voting and invoking the Charter. What has worn away is the common ground, a shared account of what is happening and of who is breaking the rules. The erosion is sharpest between Russia and everyone else, it speeds up after 2011–2014, and it is strongest where a member's own interests are on the table.
Finding 1
Vetoes became frequent. Veto votes on these conflicts went from — in 1991–2010 to — in 2011–2024. Over a three-year window, the share of decisions adopted unanimously fell from — to —.
The words parted at the same moment. Russia's statements moved away from the other four members' from the late 2000s, and the whole Council became more divided after Crimea in 2014.
Accusation followed. In a coded sample of 100 paragraphs per member and period, the share of paragraphs using a norm to accuse another actor rose for every member. Russia's went from — to —.
Dissent on contested decisions rose for all five members after 2011: —.
Votes are the firmest evidence here: every veto is a public record. Word distances rest on UN English translations for France, Russia and China, and some years have few meetings. The accusation shares come from a machine-coded sample of 3,000 paragraphs, so each period rests on 100 paragraphs per member. The dates line up, but that alone does not say which change drove the other.
Finding 2
Each small chart is one member: its lines show how far its words are from each partner's, period by period. In Russia's chart all four lines climb. It did not move away from one rival but from everybody.
The Western charts are the mirror image: the lines to the other two Western members fall while the line to Russia rises. The gap widened from both sides.
China's chart has a different shape. In the 1990s it was much closer to Russia (—) than to the West; by 2022–24 it was about as far from Russia (—) as from the Western members (—).
Votes tell the same story from another angle: Russia cast — veto votes before 2011 and — after.
Two readings of 2024
The two methods measure different things. The Council view compares the words members choose at the same meeting, so it picks up who is arguing with whom. The Atlas compares each member's average wording over a year, where China's translated statements, with their fixed formulas, set it apart. Read together: Russia is the one that broke away in argument; China has always spoken in its own register.
Russia's rise in distance is the only member-level change that is clear of the noise over three decades, and France–Russia is the pair that grew apart most in the Atlas, with the same sign under all six alternative text treatments. Similar words can still carry opposite positions.
Finding 3
After Russia's full-scale invasion of Ukraine, the Western members invoked the Charter and international law far more than ever. France went from — of its sentences in the 1990s to — in 2022–24.
Much of that is accusation. The dark part of each bar is Charter language in the same sentence as a violation word. For the United States it went from — of its Charter sentences to —.
Russia and China did not turn to the Charter. They turned to contesting how it is applied: unilateral sanctions, double standards, hypocrisy. China's use rose from — to —, Russia's from — to —.
Who denies whose legitimacy, 2011–2024: the Western members denied Russia's — times; Russia denied that of the United States and "the West" — times. Both camps claim the order; each says the other breaks it.
The dictionary was checked by hand on 1,347 matched sentences: every vocabulary scores 0.89 precision or above after fixes, except contestation in British statements (0.74). The legitimacy judgments were coded by a language model from 13,481 sentences and are not hand-checked across the board; implied objects are the least reliable. Counts per year also follow how often the Council met.
Finding 4
In the early 1990s the three Western members almost always voted alike, and Russia's words were the closest to everyone's.
By 2024 Russia and China voted the same way on — of contested decisions, and — of China's — vetoes were cast alongside Russia. Yet their words are among the furthest apart of any pair. And their judgments run one way: in 34 years of speeches Russia never judged China's actions, while China never denied Russia's legitimacy and acknowledged its "legitimate concerns" — times, mostly over Ukraine.
The Western members became the closest speakers on the Council and rarely deny each other's legitimacy (— times in 34 years). In votes they drifted: their agreement fell from — in 1991 to — in 2024. Across the two blocs it fell from — to —.
Vote agreement covers contested decisions only, over a three-year window, and some windows rest on few votes. Word distance is a measure of wording, not of policy: two members can use similar words to argue opposite positions.
Finding 5
Where a member is a party to the conflict or is shielding a close partner, it voted something other than yes — of the time, against — on all other contested decisions.
Vetoes cluster on the conflicts closest to the vetoing member: Russia cast — of its — veto votes on Syria, the United States — of its — on Israel–Palestine.
The words show the same pattern. Charter language rises most among members objecting to someone else's conduct: after Iraq 2003 for France, China and the UK, after Ukraine 2022 for the West. Members invoke the rules most when someone else is breaking them.
The own-stake flags are analytic labels with a source each, not legal findings, and China carries none because no sourced case was found. The pattern is an association: nothing here shows why a member voted as it did.
Measured by both rhetoric and behaviour, the five permanent members still treat the Security Council as the place to deal with inter-state conflicts, but their commitment to working through it together is conditional. It held while the Council's work did not touch a permanent member's core interests and weakened when it did: Kosovo in 1999, Libya and Syria after 2011, Crimea in 2014 and Ukraine in 2022. The Council has become less a room where a shared account of a war is negotiated and more a stage where competing accounts are put on the record, with both camps claiming the rules for themselves.
Read with care. French, Russian and Chinese statements are UN translations into English. The legitimacy and Lab coding is machine-made. Vote-level records cover 123 of the 181 contested decisions. Every pattern here is an association, not a cause.
UN Security Council · the five permanent members · —
How far have the five permanent members backed collective action when states go to war, and how has that changed? We follow every contested vote on violent inter-state conflicts, seat by seat.
Of — voted Security Council decisions on our conflicts, — passed unanimously. Where the members split, the line runs between Russia and China and the three Western members, and the two sides defend their votes with different ends.
Chapter 1 · Consensus
—
The dark river is the share of voted decisions adopted unanimously, over a three-year window; the pale banks are the rest. Each red line is a veto.
What changes and why: unanimity becomes one continuous river with every veto cut into it; rates use three-year windows with n shown; "adopted without a vote" no longer counts as contested.
A decision is contested when it was not unanimous or a permanent member did not vote yes. The line shows how many decisions each year passed unanimously; the bars count the contested ones.
Every voted decision on our conflicts. Hover a year for its numbers.
Chapter 2 · Cracks
—
Each column is one contested decision, in date order; each row is a member. A faint tick is a yes, hatching an abstention, solid dark a No on a draft that failed anyway, red a veto. Drag across the strip to pick a stretch of time; click a column to open the decision.
—
—
Chapter 3 · Blocs
—
Agreement is the share of contested decisions on which two members took the same position, over a three-year window. Squares under the axis mark years when the UK or France voted differently from the US.
The Council table already answers "who votes with whom" clearly, so it stays as it is.
Each seat is a permanent member. A thicker line joins two members who took the same position (yes, abstain or against) more often on contested decisions in a three-year window around the year shown. The coloured ring is the member’s dissent rate: the share of contested decisions on which it did not vote yes. Click a seat to explore that member below.
P3 = the United States, the United Kingdom and France (mean of their three pairs). The pair data are not split by conflict, so the conflict filter does not apply here.
Chapter 4 · The veto
—
Each red cut is one veto vote; a line between Russia and China joins a joint veto. Click any mark for the record.
What changes and why: vetoes become red cuts on one time axis, joint vetoes are linked, members with none keep their row, and every mark opens its decision card with record and source.
One mark per veto. A diamond is a joint veto, cast together with another permanent member on the same draft. France and the United Kingdom cast none.
Hover a mark for the draft and its subject.
—
—
—
Pick a real failed draft and switch any permanent member's vote. The outcome is recalculated from the real tally under Article 27: at least nine yes votes and no permanent member voting No.
Recalculated outcome
—
—
This is a counterfactual built on the real vote, not a prediction of how any member would have voted.
Chapter 5 · Interests
—
Darker cells mean a larger share of a member's votes on that conflict were not yes. Cells with fewer than five votes show the count, not a rate.
Flag (a): the member was a party to the conflict. Flag (b): it was shielding a close partner. Each flag has a one-line justification and a source in the data table. These are analytic labels, not legal findings: the Charter's duty to abstain covers only Chapter VI and Article 52(3) decisions.
Chapter 6 · Accountability
—
Top lane: vetoes. Bottom lane: the General Assembly meeting that followed (diamond: a veto-initiative debate; square: an emergency special session). The curve is labelled with the days in between.
Filled circle: supports. Open circle: does not. Red bars: veto votes on these conflicts, in total and after each pledge date.
Chapter 7 · Their words
The speech analysis is your team's work, shown as published and restyled only. One Version B sits beside the explorer's vote chart.
Pick a member, and a second one to compare. The charts show how often it broke from the majority, in which periods, and which arguments it used in its speeches.
— did not vote yes on — of the contested decisions it voted on and cast — vetoes. In the coded speeches, its most frequent argument when it did not vote yes was —.
Share of contested decisions on which the member did not vote yes. Other members in grey.
Contested decisions voted on, how many it did not vote yes on, and its vetoes.
Main argument of each coded speech, as a share of the member’s speeches on that side. Click an argument to read its quotes.
| Date | Vote | Quote and argument |
|---|
What changes and why: the explorer's yearly dissent line becomes a three-year window with n on every point, split into abstentions, non-veto Noes and vetoes; rates resting on fewer than five votes are hidden.
Each speech’s main argument, split by how the speaker’s country voted. Members who voted yes mostly justify their vote by what the decision is for; members who did not vote yes mostly argue about the text, sovereignty, procedure or others’ motives. Diplomacy is the end both sides claim.
Left: speeches by members who voted yes. Right: members who abstained, voted no or vetoed.
Whether the speech says the Council is the right body to act here.
How each permanent member voted at the turning points, with the main argument from its speech where that speech is in the coded sample.
Chapter 8 · Verdict
—
No single score: each column is a separate indicator, each small chart runs through the six periods from left to right. Hollow points rest on fewer than five votes.
Where a Version B changes what the original suggests, both readings are shown side by side.
Chapter 9 · Method and data
—
Your stated totals for the uploaded CSV, recomputed on the CSV as uploaded and on the cleaned file.
Every change from the original data, with its source.
Votes come from our UN Security Council P5 database, with missing votes verified by hand from UN press releases and meeting records. Arguments come from the speeches permanent members gave at meetings with a contested vote. Gemini read each speech without being told how the member voted and picked its main argument from a fixed list of nine; the vote was joined afterwards.
Because of API limits, — of the — speeches were coded (—), drawn in random order. For — of them, the quote Gemini gave as evidence was found word for word in the speech. Speeches where a member voted differently on several drafts at one meeting are left out of the side comparison.
| Period | Speeches | Coded | Share coded |
|---|
Legitimacy is an attribute that labels an object, usually an action or decision. This part analyses what actors and what actions are labelled legitimate or illegitimate among the P5.
Headline results from every judgment coded in the full corpus of P5 speeches on the eighteen conflicts, 1991–2024. All counts are lower bounds: the word filter that picked the sentences finds about a third of the judgments made (see "How much the filter misses" in the notes at the bottom of the page).
Delegitimation among the five, per 100 conflict speeches
Red: judgments in which a member denies the legitimacy of another member or of NATO / the West, as found. Dashed box: the estimated full level if the filter misses the same share everywhere (line = central estimate). The recall test suggests 2000–13 is missed more than the rest, so that period's real level is probably nearer the top of its box.
Patterns that hold across the views below
From the 3,000-paragraph Method C coding (not the full corpus): share of each member's paragraphs in which it uses at least one norm to accuse another actor. One dot is one year, and its size is the number of coded paragraphs behind it; faded dots rest on fewer than 8. Years with fewer than 3 paragraphs are left out, and the line breaks wherever the sample skips more than three years — so a lone dot means a year the sample reaches but its neighbours do not.
Every coded judgment in the full corpus, as a matrix: the row is the actor whose action is judged, the column the member that judges it. Red = legitimacy denied, green = granted; the deeper the colour the more one-sided. Actors judged fewer than 3 times in the current selection are left out. Click a cell to read the sentences.
Judgments whose object is the chosen member: who judges its actions, and whether they grant or deny legitimacy.
Each cell counts the actions the chosen member denies legitimacy or legality to. Actors condemned fewer than 3 times are left out. Click a cell, a conflict or an actor to list the actions.
The five permanent members sit in the inner circle; every other actor they judge sits on the ring outside, grouped into blocs, institutions, states and non-state actors. Each arrow runs from the member who judges to the actor judged: red = legitimacy denied, green = granted, thicker = more judgments. Drag the slider to move through the years. Hover an arrow for the count and click it to read the sentences; click an actor to show only its ties, and click it again to let go. Only the most judged outside actors of the chosen years are drawn.
Conflict inside the Council, measured as delegitimation: every judgment in which one permanent member denies the legitimacy or legality of another's action. The bars stack by the member doing the denying. Because the corpus grows roughly tenfold between the 1990s and the 2020s, the raw number rises partly because the five simply speak more — the two rates divide it by how much each member spoke, and are the fairer comparison across decades. The speech denominator counts only speeches on the eighteen conflicts of this study, matching the scope of the coded sentences. NATO and "the West" are counted by default, since members often aim at the bloc rather than at a named capital.
One line per actor whose actions are judged (the object): solid = legitimacy denied, dashed = granted, number of judgments per year. Vertical lines mark the start of the conflicts in this study. Choose the objects and the subjects (who judges); any combination adds up.
Every coded sentence, grouped by the conflict its debate belongs to. The chart shows how many judgments each issue carries year by year; pick an issue to read its sentences.
Every actor whose actions the five judge, inside or outside the P5: red bars to the left = legitimacy denied, green to the right = granted. Choose who judges and a five-year range; click a bar to read the sentences.
Full corpus: every P5 speech in the dataset, all agenda items, about 5 million words after cleaning. How often each member uses words that deny legitimacy (illegal, violation, aggression, sham, unilateral…) or grant it (legitimate, lawful, self-defence, in accordance with…).
Each member's most used denying and granting words, per 10,000 words of its speeches. Pick a year to see that year alone.
For every coded judgment, whether the sentence grounds the claim in law. "Illegal under international law", "in violation of resolution 2254" or "war crimes" count; "threatens stability", "unacceptable" or "a dangerous precedent" do not. The wider reading also counts the Charter principles invoked as norms rather than cited as law — sovereignty, territorial integrity, non-interference, aggression, occupation, impunity. The grey line is the five together; years with fewer than 10 judgments are left out, and faded dots rest on fewer than 30.
UN Security Council · how the five permanent members talk about inter-state conflicts · 1991–2024
In the 1990s, Russia's statements at the Security Council were closer to those of the other permanent members than any of theirs were (average distance —, the lowest of the five). By 2022–24 Russia was the furthest from all of them (—). The United States, the United Kingdom and France moved closer to one another, and China barely moved at all.
How to read the numbers. For every meeting, each member's statement is turned into a profile of the distinctive words it used. Cosine distance compares two profiles: 0 means identical wording, 1 means nothing distinctive in common. Diplomats always word things differently, so values sit around 0.70–0.80. What matters is the direction: higher means further apart, and a change of a few hundredths, sustained over hundreds of meetings, is large.
Each member's average distance from the other members at the same meetings, in — compared with —. The large number is the change (Welch t-test).
Only one member's position changed clearly over three decades: Russia, whose distance from the rest grew by —. The United States (—), the United Kingdom, France and China are, on average, as far from the group as they were in the 1990s. The Council did not drift apart evenly: one permanent member drifted away from the other four.
Yearly average of each member's distance from the other members at the same meetings. Dotted lines mark the five key moments. Click members to switch their lines on or off; with one left you also see its 95% confidence band. All members switches every member line on, or off when all are on, so you can show only the group lines. Group lines (P5 in black, West in dark blue, Non-West in red, each with its own dash pattern) are averages of members' distances each year: P5 average (all five), West (US, UK, France) and Non-West (Russia, China). Turn them on to see whether a member is moving towards or away from a group.
Russia's line starts at the bottom: in the 1990s it spoke more like everyone else than any other member did. It jumps in 1999, the year of the Kosovo war. From the mid-2000s it sits above the three Western members, from 2014 it stays near the top together with China, and since 2022 it is the highest. 2014 lifts every line at once: after Crimea the whole Council became more divided. The United States, the United Kingdom and France trace almost the same path as each other. China starts as the most distant member in the 1990s, comes closer during the 2000s, and returns to the top after 2014.
Change in distance for each pair between 1991–99 and 2022–24. Coloured cells are clear changes (p < 0.05): red means the pair moved apart, blue means it moved closer. Grey numbers are changes too small to be sure of.
Every red cell involves Russia. It moved away from each member, most of all from the United Kingdom (—), then France (—), the United States (—), and even China (—). Every blue cell is a pair of Western members: the US and France (—), the UK and France (—), the US and the UK (—). China's distance from the West did not change in any clear way.
Change in each member's distance from the others in the two years after each event, against the two years before, comparing meetings on the same conflict. Coloured cells are clear changes (* p < 0.05, ** p < 0.01, *** p < 0.001).
Kosovo 1999 moved only Russia (—): NATO went to war without a Council mandate, and Russia's language split from everyone else's. Crimea 2014 pushed all five members apart at once, from the United States (—) to China (—): the whole Council became more divided. Ukraine 2022 brought smaller rises for Russia and the UK. Iraq 2003 and Libya 2011, despite the anger they caused, left no clear mark on how members talk about the same conflict.
One small chart per member, all on the same scale. Each line is the distance to one partner, averaged over the six eras that begin in 1991, 1999 (Kosovo), 2003 (Iraq), 2011 (Libya), 2014 (Crimea) and 2022 (Ukraine). Line colours name the partner.
In Russia's chart all four lines climb from the 2000s onwards: it moved away from everybody, not from one rival. In the US, UK and French charts the same thing happens in mirror image: the lines to the other two Western members fall while the line to Russia rises, so the gap between "the West" and Russia widened from both sides. China's chart has a different shape. In the 1990s it was much closer to Russia than to any Western member; today it is about as far from Russia as from the West. The "Russia–China axis" seen in votes does not show up as a shared way of talking about conflicts.
Mean distance for each of the ten pairs. Darker is further apart. Rows are grouped: Western pairs, then each Western member with Russia, then with China, then Russia with China.
Read across a row to follow one relationship. The Western pairs at the top stay the lightest. The Russia rows darken from 2014 onwards and stay dark. Single dark cells in early years usually rest on only a handful of meetings; hover to see how many.
Each member's average distance from the others, by how the meeting ended, 1992 to September 2022. Rows are sorted from the most to the least divided.
For every member, meetings that ended in a veto or a failed draft show the largest distances. For Russia that is — at veto meetings. Presidential statements, which need every member's agreement, show the smallest (—). So the measure is not just noise: when the members' words part, the Council tends not to act. Part of this gap reflects when and on which conflicts vetoes happen.
Before the Council can act on a conflict, its members must describe it in compatible terms: what happened, who is responsible, which rules apply, what should happen next. These charts measure how far apart those descriptions are. They do not measure trust directly. They show whether the permanent members still treat the Council as a room where they work from a common account of a crisis.
Right after the Cold War, Russia's statements were the closest of any member to the others' (—). That fits the post-Cold War hope that the Council could finally work as the UN Charter intended, with the great powers addressing conflicts together.
The first came with Kosovo in 1999, when NATO acted without the Council's authorisation, a direct challenge to the Council's role that Russia contested. The second came with Crimea in 2014, when Russia itself became a party to a conflict on the agenda. By 2022–24 Russia was the furthest from every other member, China included.
The United States, the United Kingdom and France moved steadily closer to one another. They keep using the Council as their main forum and increasingly speak there as one bloc. The other side of that coin is that Council debates have become more two-sided.
China's distance from the group did not change (—). It is no further from the West than in the 1990s, but it has moved away from Russia. China and Russia often vote together, yet they do not describe conflicts in the same way. China stays engaged on its own terms rather than joining either camp.
The largest distances come at meetings that end in a veto or a failed draft, the smallest at presidential statements, which need consensus. The drift in language goes hand in hand with the Council's growing inability to act together.
The permanent members have not walked away from the Security Council. Russia made statements at — meetings on these conflicts in 2022–24, against — over the whole of 1991–99. What has eroded is the common ground. The Council is increasingly a stage where competing accounts of a war are put on the record, rather than a room where a shared account is negotiated. That erosion is concentrated in Russia's relationship with everyone else, and it sped up when Russia's own interests were on the table. This fits a reading in which great-power commitment to the multilateral order is conditional: strong while the Council's work does not touch a permanent member's core interests, weak once it does.
The Western members show the opposite tendency. Their commitment shows up as closer coordination with each other, which keeps the Council in use as a forum but turns it into a place of bloc positions. China signals continued engagement without alignment. For the research question, cosine distance captures the rhetorical side of commitment. It should be read together with the voting and veto record, the behavioural side.
UN Security Council · what the five permanent members say about the rules · 1991–2024
When the P5 speak about a conflict, how often do they invoke the UN Charter, the Council's authority, sovereignty or the ban on force, and how often do they contest the order itself? In the 1990s France used the language of the Charter and international law in — of its sentences. By 2022–24 that had risen to —, with the UK and the US close behind, while Russia (—) barely changed. Meanwhile Russia and China turned to contestation: China's talk of unilateral sanctions, double standards and hypocrisy rose from — to —.
How to read the numbers. Every speech is split into sentences. A sentence counts for a vocabulary when it contains at least one of that vocabulary's terms from a validated dictionary (for example "territorial integrity" or "double standards"). A speech's score is the share of its sentences that use the vocabulary. Members' scores are averages over their speeches, so long and short speeches count equally. Values are small because most sentences of a speech are about facts on the ground: 1% means about one sentence in a hundred, and changes of one or two points are large. Choose a vocabulary in the bar below; every chart on this page follows it.
Each member's average share of sentences using the chosen vocabulary in 1991–1998, compared with 2022–2024. The large number is the change in percentage points (Welch test).
Average share of sentences per speech, by member and year. Dotted lines mark the five key moments. Click members to switch their lines on or off. With one member left and Yearly selected, the shaded band is the 95% confidence interval (bootstrap over speeches). 3-year average smooths out years with few speeches.
Sentences using the vocabulary, split by whether they also contain a violation word ("violate", "breach", "flout", "undermine" and similar). The dark part of each bar is accusation, the light part is everything else. Top bar 1991–98, bottom bar 2022–24.
Change in each member's use of the chosen vocabulary in the two years after each event, against the two years before (all conflicts in the sample). Coloured cells are clear changes (* p < 0.05, ** p < 0.01, *** p < 0.001): red means used more after the event, blue less.
One small grid per member: rows are the seven vocabularies, columns the six eras that begin in 1991, 1999 (Kosovo), 2003 (Iraq), 2011 (Libya), 2014 (Crimea) and 2022 (Ukraine). Shading is comparable across members within a row: the darkest cell in each row is the highest value any member reached for that vocabulary. The chosen vocabulary is outlined.
Read down a column to see what a member stressed in one era, and across a row to follow one vocabulary. Two rows fade for everyone: Council authority and peaceful settlement were most present in the 1990s. The use of force row darkens for France and the UK in 2022–24. The US row was already dark in the 1990s, the years of the Gulf War and its aftermath. The contestation row darkens steadily for Russia and, after 2014, for China. China's grid is the darkest overall: its translated statements lean heavily on the formulas of the Charter, sovereignty and settlement, so compare China with its own past rather than with the others.
Average share of sentences using the chosen vocabulary, by conflict and member, over all years. Darker is more frequent. Conflicts are sorted by number of speeches; hover for counts. Cells with a dashed outline rest on fewer than 20 speeches and are not shaded. Comparing members within a row removes the effect of which conflicts were on the agenda.
The ten most frequent terms of the chosen vocabulary, and how often each member uses them (sentences per 1,000 sentences). A row where the dots spread far apart is a term that one member owns.
Real sentences from the validation sample that the dictionary counted for the chosen vocabulary, each checked by hand as a correct match. Choose a member; Other examples shows a different set.
Precision: the share of matched sentences that really use the vocabulary, from a stratified sample of — matched sentences coded one by one. Light bars show the first draft of the dictionary, dark bars the final version after weak terms were fixed. The dashed line marks 0.8, a common threshold.
Matches for the Charter, peaceful settlement and sovereignty were right almost every time from the start. Four vocabularies needed fixing. "Primary responsibility" often meant a government's responsibility rather than the Council's, "collective security" was often the name of a regional alliance, "use of force" often described policing or proportionality, and "obstruct" often meant blocked aid convoys. After those terms were restricted, every vocabulary is at 0.89 or above. The weakest spot is contestation in British statements (—). A team member checked a random subset of the coding and agreed with it 89% of the time. What the dictionary misses (recall) is not measured yet.
Talk of the Council's own authority, its unity and multilateral action fell for all five members (clearly for four of them), most steeply for China (— → —). So did calls for peaceful settlement. Thirty years after the Cold War, the P5 spend less of their speaking time on the Council as the place where conflicts get solved.
With Russia's full-scale invasion of Ukraine, France, the UK and the US invoked the Charter and the ban on force far more than ever before. France's use of force vocabulary went from — to —. Much of this new Charter talk accuses Russia of breaking the rules: in 2022–24 nearly a third of France's Charter sentences contain a violation word, against about one in ten in the 1990s.
Russia's and China's Charter language did not grow. What grew is contestation: unilateral sanctions, double standards, politicisation, hypocrisy. Russia's rose from — to —. They rarely reject the Charter itself. Instead they accuse others of applying it selectively, a way of claiming the order while denying its current custodians.
Charter language rises around specific events, and it rises among those who object. After Iraq 2003, France, China and the UK used more of it; after Ukraine 2022, the West used far more. Members invoke the rules most when someone else is the one breaking them. This fits the conditional commitment seen in the other views.
In every era China's statements contain the most Charter, sovereignty and settlement vocabulary. Part of this is the style of its translated statements, which repeat fixed formulas. China's own trend still shows sovereignty and Council-authority talk falling since the 1990s, and contestation rising sharply after 2018.
Both camps invoke the order, for opposite ends. For the West it is a charge sheet against Russia. For Russia and China it is a standard the West fails to meet. This matches the drift measured in the Council view: the members still share a vocabulary of rules but no longer a common account of who is breaking them.
Measured by what they say, none of the P5 has abandoned the UN-backed order: all five still invoke the Charter, sovereignty and international law at every stage. What changed is how they use that vocabulary. The language of joint work through the Council and of peaceful settlement has faded. In its place, the rules have become ammunition: the Western members use them to indict a permanent member that invaded its neighbour, while Russia and China use the language of double standards and unilateralism to contest the West's authority to enforce them. Commitment to the order now shows up mainly as competing claims to own it.
These are counts of words, not measures of sincerity or behaviour. They should be read alongside the voting and veto record and alongside the distance measures in the other views.
Every P5 speech, meeting record, decision and veto, joined by explicit keys. Search the texts, read the automated C1–C7 labels, compare members as parties to a conflict and trace what the Council decided. Click any count or mark to open its records.
Loading the research data…
UN Security Council · P5 statements on inter-state conflicts · 1991–2024
Where the five permanent members stand in a shared space of diplomatic language, year by year. The map shows who and where, the semantic plane shows how close their words are, the lexical axes show why, and the timeline shows when.