The Blank Space in the Treaty: Why Taiwan's Status Was Never Written Down

In September 1951, forty-eight states signed a peace treaty with Japan in San Francisco. Article 2(b) says Japan renounces all right, title and claim to Formosa and the Pescadores. It does not say to whom.

That omission wasn't an oversight. By 1951 there were two governments each claiming to be China, a war on in Korea, and no agreement among the signatories about which of the two should receive anything. Neither the People's Republic nor the Republic of China was a party to the treaty. So the drafters renounced without transferring, and left the sentence to end early.

Seventy-five years later, most arguments about Taiwan are still arguments about how to fill in that blank.

What the documents actually did

The chain of texts is short and each link is weaker than people assume.

Japan took Taiwan and Penghu from Qing China by the Treaty of Shimonoseki in 1895 — a real cession, in a real treaty, and the only unambiguous territorial transfer in the whole sequence. Fifty years of colonial rule followed. It built railways, schools and public health systems, and it was coercive, unequal colonial government at the same time; both things are in the record.

The 1943 Cairo Declaration stated that territories Japan had taken from China, Formosa and the Pescadores among them, should be restored to the Republic of China. Potsdam in 1945 called for Cairo's terms to be carried out, and Japan's surrender accepted Potsdam. ROC authorities began administering Taiwan that year.

Then came 1947, and the February 28 uprising, and its violent suppression, and the decades of White Terror and martial law that followed. Taiwan's later democracy grew up in opposition to two things at once: the Communist state across the water, and the authoritarian ROC state on the island itself.

The Communists won the mainland in 1949. The ROC government moved to Taiwan with its army, its officials and a wave of mainland civilians. No armistice ended the cross-strait phase of that war. There is still no peace treaty.

Which brings us back to San Francisco in 1951, and to the Treaty of Taipei the following year, in which Japan and the ROC acknowledged the renunciation — and whose operative text, again, contains no explicit transfer of the territory to the ROC.

Four readings of the same blank

Beijing's position is one of state succession: Taiwan was restored to China in 1945, the PRC succeeded the ROC as China's sole government, and sovereignty came with the succession. This explains Chinese state-continuity claims and Beijing's hostility to anything resembling "two Chinas." It's harder to explain how succession delivers sovereignty over territory the PRC has never administered for a single day.

The older KMT reading runs through Cairo, Potsdam, the surrender, ROC administration from 1945 and the Treaty of Taipei, and concludes that Taiwan was restored to the ROC. That explains why Taipei has treated the island as its own territory for eighty years.

Its best textual hook is Article X of the 1952 treaty, which deems the inhabitants of Taiwan and Penghu to be nationals of the Republic of China. That's the closest any of these documents comes to naming a recipient — and it names people, not territory. What the reading can't establish is whether wartime declarations and surrender arrangements finished a transfer the peace treaty itself declined to write.

The unsettled-status reading just points at the 1951 wording and stops there. Japan gave up title; nobody was named; the treaties say what they say.

And there's the self-determination argument, which sets the older title aside and says Taiwan's future belongs to the people who live there now. Beijing rejects any process that could make separation permanent, and international law doesn't hand down an automatic answer to that, so this reading is a political claim about legitimacy rather than a legal trump card.

None of this touches who actually governs. Taipei does. Beijing never has. That question and the sovereignty question have different answers, and only one of them is disputed.

The resolution that doesn't mention Taiwan

UN General Assembly Resolution 2758, adopted in 1971, recognised the PRC's representatives as the only legitimate representatives of China at the UN and expelled the representatives of Chiang Kai-shek. It settles one question: who sits in China's seat.

Taiwan isn't mentioned in it anywhere — not as territory, not as a population, not even as a question the Assembly was choosing to leave for later.

Beijing's argument is a syllogism laid over the text — the PRC is China's sole representative, Taiwan is part of China, therefore 2758 also forecloses separate Taiwanese representation. Washington and a growing number of other governments now argue publicly that this stretches the resolution past what it says, and a 2025 Congressional Research Service review separates the representation decision from the unresolved status dispute in exactly those terms.

The argument is legally interesting and practically almost irrelevant. Taiwan is excluded from the UN system and most of its specialised agencies regardless, because Beijing has the diplomatic weight to keep it out. Taipei participates in the WTO and APEC under negotiated names, which is roughly the ceiling.

One verb, holding up the whole arrangement

The 1972 Shanghai Communiqué says the United States acknowledged that Chinese on both sides of the Strait maintained there was one China and Taiwan was part of it. In 1979 Washington recognised the PRC as the sole legal government of China, and again acknowledged the Chinese position on Taiwan.

Acknowledging that someone holds a position is not adopting it. Fifty years of American Taiwan policy sit on that distinction.

So there are three "one China" formulations, and collapsing them produces most of the bad commentary on this subject. Beijing's one-China principle is a sovereignty claim and a precondition for diplomatic relations. Washington's one-China policy is a constructed framework: recognise the PRC government, acknowledge its position, keep unofficial ties with Taipei, oppose unilateral change, support peaceful resolution. Taiwan's own positions are plural by design, because it's a democracy — they run from ROC constitutional continuity and "one China, respective interpretations" through to the view that Taiwan is already sovereign and simply happens to be called the ROC.

The Taiwan Relations Act of 1979 rewards the same close reading. It preserved unofficial relations, authorised defensive arms, and called efforts to decide Taiwan's future by other than peaceful means — boycotts and embargoes included — a grave concern to the United States. On the question everyone actually cares about, it declares. It does not oblige. Maintaining the capacity to resist coercion is stated as American policy, and Section 3(c) hands any actual response to the President and Congress "in accordance with constitutional processes."

That is the arrangement working as designed. Beijing has to weigh an intervention that might come. Taipei has to plan around one that might not, which is a constraint on Taipei as much as on anyone.

What the ambiguity is doing now

Ambiguity was supposed to buy time. It has, so far. It has also created a long, finely graded ladder for applying pressure without anybody having to call it war.

Taiwan's 2025 National Defense Report describes the middle of that ladder without putting a number on it. Median-line crossings have stopped being exceptional. Warships work the 24-nautical-mile line. Balloons drift over. Exercises rehearse blockade and assault rather than gesturing at them.

Underneath sit the things that never look like war at all: cyber intrusion, information operations, diplomatic isolation, economic pressure. Further up, a coast-guard "quarantine" of selected shipping, then a real blockade, then limited strikes or the seizure of an offshore island. An amphibious invasion is the top rung and the least likely one.

The quarantine rung is the interesting one. Framed as law enforcement inside territory Beijing already claims, it would put the burden of the first escalation on somebody else: on Taipei, on Washington, on a Danish shipping line deciding whether to sail. It exploits an island dependent on maritime trade and imported energy without requiring anyone to storm a beach.

Invasion is genuinely hard. Roughly 130 kilometres of water at the narrowest point, air and maritime superiority to establish first, ports or beaches to seize, sustained logistics, urban combat, and 23 million people to control afterwards, with possible US and Japanese intervention throughout. The 2026 US Annual Threat Assessment says PLA progress is steady but uneven, that Chinese leaders have no fixed unification timeline and do not currently plan an invasion in 2027, and that Chinese officials themselves recognise the operation could fail.

On which: 2027 is a PLA modernisation and readiness benchmark. It has been repeated so often as a deadline that it now functions as one in public debate. Capability, intent, and a decision to act are three separate things, and only the first of them has a date attached.

The island that voted anyway

In 1995 and 1996, after President Lee Teng-hui visited the United States, the PLA fired missiles and ran exercises near Taiwan. Washington sent two carrier groups. Taiwan held its first direct presidential election on schedule in March 1996.

Intimidation aimed at Taiwanese political identity has a record of strengthening it. Martial law ran from 1949 to 1987; the press opened, opposition parties became legal, the old mainland-elected legislature was retired, and power passed peacefully from the KMT to the DPP in 2000. By the June 2026 reading of National Chengchi University's long-running survey, 63.1% of respondents identified as Taiwanese only, 30.7% as both Taiwanese and Chinese, and 2.5% as Chinese only.

Identity is not the same question as constitutional status, and the polling is careful about the difference. Asked about status rather than identity, broad versions of "maintain the status quo" consistently take the largest combined share. The 2024 election gave Lai Ching-te the presidency on 40.05% in a three-way race, and gave nobody a legislative majority: 52 seats to the KMT, 51 to the DPP, 8 to the TPP, 2 independents, with 57 needed. Recall votes in 2025 left that arithmetic intact.

An election isn't a sovereignty referendum. A DPP vote can be about identity, or security, or healthcare, or the candidate; a KMT vote can be about the economy or institutional checks or a preference for talking, without implying any enthusiasm for rule from Beijing.

The shield, and its wiring

Taiwan's other form of protection is industrial. TSMC's 2025 annual report puts processes at 7 nanometres and below at 74% of wafer revenue, with 2-nanometre high-volume manufacturing beginning in late 2025. War would impose enormous costs on China, the United States and everyone else, which gives most major economies a direct stake in Taiwan not being attacked.

Leaders have accepted catastrophic economic losses for political objectives before. And a fab is not a fortress — it runs on electricity, water, imported materials, foreign software and international customers, all of which a blockade could sever while leaving every building standing. TSMC's expansion into Arizona, Japan and Germany reduces concentration risk over time; it doesn't reproduce a cluster built out of engineers, suppliers, utilities and accumulated process knowledge.

Taiwan's trade map is shifting too. The Ministry of Finance's annual external trade report for 2025 shows the United States overtaking China and Hong Kong as Taiwan's largest export destination — about 30.9% versus 26.6%, against 31.7% for China and Hong Kong in 2024 and 43.9% in 2020.

Meanwhile Taipei has raised 2026 defence spending to a reported NT$949.5 billion, or 3.32% of GDP under a broad NATO-style accounting that folds in veterans and coast-guard costs, and extended conscription from four months to one year starting in 2024. The harder problems aren't budgetary. They're mobilisation, realistic training, reserve organisation, stockpiles, command continuity, and how quickly purchased weapons actually arrive.

Six questions, not one

As of 4 September 2026, twelve states maintain formal diplomatic relations with Taipei. That number tells you almost nothing on its own, since the United States, Japan, most of Europe, India and Australia all run representative offices and substantial commercial and security relationships without it.

Which is the recurring shape of this whole subject. Control, recognition, representation, sovereignty, identity and strategic commitment sound like facets of one question and behave like six separate ones. Taipei has control and almost no recognition. The PRC holds the UN seat while never having administered the island. Identity has moved decisively toward Taiwanese without settling anything constitutional, and American commitment stays unstated on purpose.

Take the UN seat as an answer about sovereignty, or effective control as an answer about recognition, and you arrive at a confident conclusion the documents don't carry.

The peace that has held since 1949 rests on three restraints, none of them written into any treaty: Taipei doesn't make the formal constitutional declaration that would force the question, Beijing prefers unification without a war while building the capacity to fight one, and Washington arms Taiwan without promising to defend it. It's an arrangement built on a sentence that stopped before naming anyone, and it has lasted longer than most of the settlements that named names.

Sources: Treaty of Shimonoseki, 1895 · San Francisco Peace Treaty, 1951 · Treaty of Taipei, 1952 · UNGA Resolution 2758, 1971 · CRS review of Resolution 2758, 2025 · 1972 Shanghai Communiqué · 1979 normalization communiqué · Taiwan Relations Act · Six Assurances, 1982 · Strait crises of 1954–55 and 1958 · PRC 2022 white paper · 2025 National Defense Report · US Annual Threat Assessment 2026 · 2024 election results · NCCU identity survey · TSMC 2025 letter to shareholders · conscription extended to one year · MOFA list of diplomatic allies · MOF external trade report 2025 · CNA on the 2026 defence budget

Four notes for anyone following the links. Both UNTS volumes are scans, so a text search finds nothing and you have to read the page. The State Department crisis page stops in 1979 and covers neither 1995 nor 1996. The NCCU numbers exist only inside the chart image, not in the page text, and the 74% sits in chapter 1 of TSMC's annual report rather than on the landing page. The defence budget figure comes from press reporting of the premier's announcement rather than a budget document, and dni.gov turns away enough readers that the Threat Assessment link goes to a Wayback capture.