What Allies Want: Reconsidering Loyalty, Reliability, and Alliance Interdependence

Do states judge their ally's behavior toward its other allies? If yes, how? Historically, decisionmakers have instinctively adopted deterrence theory's logic that a state's character is judged through displays of innate loyalty: if a state is disloyal to one ally, then this will create a reputation for disloyalty, which will cause other allies to doubt the state's reliability. Thus, disloyalty can have calamitous consequences: the aggrieved ally will punish the betrayal; other allies will suffer crises of faith; and adversaries will conclude that the state's alliances are cheap talk. The logic is that discrete alliance commitments are interdependent—that what happens in one alliance affects the expectations of other allies—and that this interdependence is underpinned by demonstrations of loyalty. President Lyndon Johnson said that if the United States were “driven from the field in Viet-nam, then no nation can ever again have the same confidence in American promise or in American protection.”¹

These convictions regularly animate contemporary debates: Nancy Bernkopf Tucker and Bonnie Glaser argue that if the United States were to abandon Taiwan in a conflict with China, this could deal “a fatal blow to the U.S.-Japan alliance” and might lead to South Korea “renouncing its security alliance with Washington and aligning with Beijing.”² Aaron Friedberg writes that expecting a U.S. “back down … [and] a Chinese victory over Taiwan … to leave America's Asian alliances unscathed, is to indulge in wishful thinking of the most dangerous kind.”³ Others suggest cross-regional effects, claiming that by “retreating from the Middle East and abandoning Ukraine to Russian aggression,” President Barack Obama left “America's Asian allies … bewildered and alienated.”⁴ If alliance interdependence exists, and is governed by innate loyalty, then fighting for reputation is crucial, because any single alliance rift could quickly tear asunder other alliance relationships.

Some scholars, described as “reputation skeptics,” dispute this common wisdom. Skeptics argue that because “reputation is in the eye of the beholder,” the United States should never regard demonstrating loyalty as sufficient grounds for military action.⁵ Jonathan Mercer argues that when allies observe the United States demonstrating loyalty, they will attribute this desired behavior to situational causes and thus will not conclude that it will be loyal in future crises. In contrast, he concludes that when the United States is disloyal, this undesired behavior will be attributed to national character, but will not always cause allies to expect similar behavior in the future.⁶ Reputation skeptics believe that “leaders are tragically mistaken when they commit to the use of force in the expectation of long-term benefits beyond any gains in the immediate dispute.”⁷

This interdependence debate has usually focused on adversarial interactions and reputations for resolve, defined as “the extent to which a state will risk war to keep its promises and uphold its threats.” Loyalty is usually assumed to be the obvious “flip side” of resolve, and is not often studied in its own right.⁸ Some studies of reputation and alliances exist, but are predominantly focused on alliance formation rather than management.⁹ Because loyalty has not received the significant scholarly attention centered on resolve, alliance interdependence is undertheorized.¹⁰ But as the contemporary debate about Taiwan highlights, beliefs about loyalty and alliance interdependence could be a decisive influence on policy today. If alliance interdependence exists, scholars and policymakers urgently need to know how it works.

I argue that both the deterrence theorists and the reputation skeptics are wrong. Alliance interdependence exists—what happens in one alliance can affect others—but this is not determined by a national moral character of innate loyalty. When a state observes its ally's behavior in another alliance, it does not hope to see loyalty per se. Instead, it hopes to see proof that the ally's interests align with its own. In some cases, it will be in the observer state's interest for its ally to fulfill a separate alliance commitment—that is, the state will want its ally to be loyal to the other ally. But in other cases, where loyalty would not be in the observer state's interest, it will want its ally to be disloyal to the other ally. This reasoning contrasts starkly with other theories, which expect that a state will always want its ally to demonstrate inherent, pure loyalty. I argue that the state is unconcerned about whether its ally's behavior is loyal or disloyal in any objective or moral sense. Rather, when the state is confident that its ally's interests align with its own, and therefore the alliance poses no risk of either abandonment or entrapment, then the ally is reliable.¹¹ Observed reliability—not innate loyalty—is what states want from allies. If an ally poses risks of entrapment, abandonment, or both, then it is unreliable, and the state will act to mitigate these risks. Responses could include hedging, diplomatic restraint initiatives, intra-alliance bargaining, threats of defection, attempts to entrap the ally, or, in extremis, alliance abrogation, bandwagoning, or nuclear proliferation.¹²

Deterrence theory recommends that states fight for reputation, and skeptics urge policymakers not to worry about reputation costs. I argue that adroit alliance management requires understanding how developments in one alliance can affect the reliability perceptions and behavior of other allies. The United States may need to fight—or not fight—to ensure that allies still perceive it as reliable. If the true nature of interdependence is recognized, the United States can manage alliance interdependence without being captive to a “cult of reputation.”¹³

In the first section of this article, I examine existing theories of reputation, loyalty, and alliance interdependence, and critique the belief that states want their allies to be indiscriminately loyal. In the second section, I explain why the concept of alliance reliability has greater explanatory power than the idea of loyalty. I propose hypotheses on reliability and alliance interdependence and contrast these with deterrence theory and Mercer's reputation-skeptic argument. In the third section, I use process tracing and archival research of the First Taiwan Strait Crisis (1954–55) to test my theory of alliance reliability. I show that although U.S. decisionmakers fretted about being perceived as disloyal to the Republic of China (ROC, or Nationalist China), most allies actively encouraged the United States to compel an ROC withdrawal from disputed territories and thus reduce the risk of conflict. In the final section, I summarize my findings and explain their implications for theory and policy.

Reputation, which Mercer states is “a judgement of … character (or disposition) that is then used to predict or explain future behavior,” is usually presented as a necessary condition for interdependence.¹⁴ Deterrence theorists argue that issuing credible threats to adversaries requires a reputation for resolve. Glenn Snyder and Paul Diesing replicated this logic by arguing that alliance interdependence is determined by a reputation for loyalty: just “as ‘resolve credit’ with adversaries can be earned and ‘banked’ by repeated instances of firmness, so ‘loyalty credit’ with present or potential allies can be generated and drawn upon in the future by repeated demonstrations of support.”¹⁵ Thus, an instance of disloyalty not only risks “the defection of an alienated partner,” but contributes toward a wider reputation for disloyalty.¹⁶

Given its importance, the concept of loyalty receives insufficient attention in the literature on alliances.¹⁷ Does loyalty mean always supporting an ally, or just complying with an alliance treaty? Is it disloyal to abandon a reckless and provocative ally? When discussing interdependence, alliance theorists are largely silent on these questions and instead transpose deterrence theory's logic onto alliance commitments: states observing an ally's behavior make a dispositional (character) judgment based on a state's loyal or disloyal conduct.¹⁸ According to this logic, because an ally's conduct reveals its innate national character and suggests similar behavior in future scenarios, states will never want to see their ally demonstrate disloyalty.

Mercer—the most prominent reputation skeptic—argues that when states observe an ally's behavior, they do not always attribute it to character.¹⁹ He claims that an ally's desired behavior will be attributed to situational factors, but its undesired behavior will be blamed on national character. The result is that “states get no credit when they behave properly, but do get blame when they misbehave.”²⁰ Thus, states should never fight to demonstrate loyalty, because this will not create a beneficial reputation. Mercer extends this argument to conclude that a state should be unconcerned about having a reputation for disloyalty, because deliberate efforts to demonstrate loyalty will be discounted by allies who will “assume that our capabilities and interests determine our resolve.”²¹ These different positions lead to diametrically opposed policy recommendations: depending on who is asked, fighting for reputation is either essential or folly.

In this debate, alliance interdependence is usually considered analogous to the interdependence of threats issued to adversaries.²² But as Gregory Miller writes, “Scholars … often ignore … a state's willingness to honor its [alliance] promises.”²³ This neglect has resulted in two contestable ideas being taken as true.

The first is that states always want to see their ally demonstrate loyalty: a term often used, but rarely defined. If Mercer's definition of resolve was truncated, loyalty could be defined as “the extent to which a state will risk war to keep its promises.”²⁴ This is the most common meaning of the term, and it directly connects loyalty to the risk of abandonment: a disloyal state does not keep its promise to fight, and thus abandons its ally. This definition, however, does not adequately encapsulate the cooperative efforts expected—and obligations felt—within alliances. Also, it suggests that judging a state's loyalty is a straightforward, unproblematic task. In contrast, Snyder argues that alliances create a norm that the states “ought to give each other mutual support on lesser issues, most especially those that relate somehow to the ultimate military contingency.” He calls this the “alliance halo,” which creates “an obligation to support, or at least to avoid damaging, the interests of the ally.” Snyder concludes that “there appears to be a norm … that allies’ expectations of support must be satisfied, even when such expectations are not founded on any formal obligation.” Thus, I borrow from Snyder and define loyalty as fulfilling not only the obligation to support an ally in wartime, as usually required by an alliance agreement, but also fulfilling the obligation to support the ally within the expectations established by the alliance halo.²⁵

Significantly, Snyder's halo concept suggests that there are degrees of alliance loyalty, a notion that confounds the description of a state as either categorically loyal or disloyal. Ambiguity in an alliance text, or differing interpretations, can make assessments of loyalty difficult: a state might be loyal on the core issues clearly formalized by an alliance treaty, but disloyal on some peripheral issue.²⁶ Thus, it is more accurate to describe specific state behavior—rather than the state itself—as loyal or disloyal.

Theorists and decisionmakers overlook these nuances and instead assume that states want their ally to demonstrate a national character trait of unalloyed loyalty. On closer examination, however, this is a questionable belief of fragile logic. Consider, for example, a state with two allies, that come under simultaneous attack. The state's military resources are such that it can support only one ally: this ally will survive, but the abandoned ally will be conquered. Each ally will desperately desire the state's loyalty and will be unconcerned that this requires the state's disloyalty to the other ally.²⁷ A similar dynamic is described by Thomas Christensen: European allies initially welcomed the United States’ decision to defend South Korea, but later worried that the Korean War would result in the Americans either “expending their power in an unimportant area or … triggering World War III.”²⁸

Other circumstances could also prompt demands for disloyalty. A state might have two allies, and the first ally's territorial dispute risks escalating into a major war. If the second ally fears escalation more than the risks of deescalation, it will worry that the state could be too loyal and will instead want the state to be disloyal to the first ally. These hypotheticals show that a state will not always want its ally to loyally fulfill all alliance commitments.

Another questionable belief is that states observe behavior to assess national character, rather than interests and capability. This assumption is strongest in deterrence theory, but even Mercer expects that reputations, based on a nation's character, might form as a result of undesired behavior. Although at one point he argues that states “usually prefer … allies to stand firm,” Mercer goes on to depict an ally's loyal behavior as desired and disloyalty as undesired.²⁹ But if states discourage or attempt to prevent loyal behavior, these actions suggest that alliance interdependence is determined by something other than judgments of national character.

If states do not want innately loyal allies, then what do they want? Gregory Miller's work suggests that states want allied reliability. Although he does not explicitly define reliability, Miller notes that a state could be considered unreliable if it “fails to fight when obligated to do so” (i.e., abandons its ally) or “drags its ally into an unwanted conflict” (i.e., entraps its ally).³⁰ He finds that reliable states maximize their choice of allies and autonomy, whereas unreliable states have fewer alliance prospects and are more constrained by alliance treaty texts. Miller's work contributes significantly to the scholarship on alliance interdependence, but its contemporary relevance is limited by the hypotheses tested and case studies examined: such findings may not be generalizable to contemporary conditions.³¹

Miller's underlying logic suggests that in a perfect alliance, the two states would always agree on goals, resources, and methods. Because this agreement would neutralize any risk of disloyalty, neither state would fear abandonment. Just as importantly, total agreement would neutralize any risk of adventurism, so neither state would fear entrapment. This perfect alliance would be all benefit and no risk: because of their identical interests, each state would be a completely reliable ally to the other. Although it is an abstraction, this ideal of pure reliability is what allies want.

To be clear, this is not how all scholars have defined reliability: others have used the word to describe states that fight alongside their allies in wartime.³² My definition includes this aspect but also builds on Miller's earlier work: I define reliability as the degree to which allies agree on the relative value of particular interests and the manner in which the interests should be pursued.³³ Thus, the most obvious sign of a reliable ally is that it does not pose a risk of abandonment or entrapment. By contrast, an unreliable ally would be too timid or too aggressive. This does not mean that two allies need agree on every issue: allies might be reliable on one matter when their interests converge, but unreliable on another when interests diverge. Further, my concept is not restricted to wartime or moments of security crisis, but extends to peacetime cases as well.

To assess reliability, states can monitor their ally's behavior for actions that demonstrate convergent interests and capabilities that can be used to achieve those interests. This dynamic differs from the argument that observers are trying to discern a state's level of resolve, which is private information about its preference ordering.³⁴ A state's preference ordering is actually not private information, but rather a “known unknown” about which the state, and its allies, have certain beliefs.³⁵ As Frank Harvey and John Mitton argue, “Interests are not always obvious … willingness to use force is never unequivocally clear or self-evident.”³⁶ Leaders may genuinely believe that they will stand firm in a crisis, but at the moment of truth, they may instead back down.³⁷ This uncertainty is why I expect even allies, which share secret information about military capabilities and strategic interests, will closely monitor each other's actions. A state's actions not only confirm the existence of particular interests, but also reveal how these are prioritized.

There are three reasons why this idea of reliability is superior to the concept of loyalty. First, reliability has twice as much explanatory power: states worry not just about allied disloyalty (abandonment), but also about being dragged into an unwanted conflict (entrapment). The idea of reliability incorporates both risks: an ally can be unreliable because it is too timid or too aggressive.

Second, the reliability concept acknowledges that a state's behavior may not be perceived in a uniform manner by all its allies. The loyalty concept implies that there is an objective standard against which national behavior is judged: an ally can be loyal or disloyal. A state's loyalty is not so easily judged, however: its actions might be simultaneously loyal and disloyal. I argue that because states observe their ally's behavior to discern interests and capabilities —not to make character judgments—a state will not have a reputation for loyalty or disloyalty. And because different states have different interests, it is possible for two nations, each allied to the same state, to hold opposing views about that state's reliability. Reliability is thus a relative concept, contingent on a state's interests and its ally's behavior. Loyalty and reliability are not synonymous: because states have different interests, one state's disloyal ally can be another state's reliable ally.³⁸

As an example: a state has three bilateral alliances, and its third ally initiates a conflict. The first ally might encourage the state to withhold military support, because it does not desire war and believes that the third ally behaved recklessly. The second ally, which feels especially threatened by the adversary, might hope that the state is loyal and therefore helps the third ally destroy the adversary's military capabilities. If the state intervenes, the first ally will view it as unreliable, because it poses risks of entrapment. The second ally, however, pleased by the decision, will view the state as reliable. All involved have learned valuable information about the state's preference ordering: the consequences of defending the third ally were prioritized over the costs of the conflict and the damage to the first ally's reliability perceptions.

This hypothetical situation illustrates the shortcomings of the loyalty concept: if one believed that states always want their ally to demonstrate loyalty, then one would incorrectly expect the first ally to be pleased. By contrast, the idea of reliability can easily explain these varied reactions. Because assessments of reliability are not judgments of character, and two states can reach opposing conclusions about the same behavior, I argue that states do not have reputations as defined by either Mercer or Miller.³⁹

Third, the reliability idea allows for alliance commitments to be interdependent, but in a broader sense than usually expected. Mercer defines interdependence as “using past behavior to predict or explain future behavior,” and interdependence is often examined in terms of iterative crises: the state backed down in the first crisis, so it is expected to back down in a second crisis.⁴⁰ Interdependence could also operate on a much quicker schedule, however, in non-conflict situations, and with reference to revelations of interests rather than character. If a state observes something indicating allied unreliability, then it should try to mitigate this risk. A state fearing abandonment could increase its defense budget, seek new allies, peacefully settle old scores, or discuss these fears with the ally. A state fearing entrapment could employ distancing strategies, threaten to withhold support, launch peace initiatives, or even abrogate the alliance. Alternatively, because the state is not judging character, interdependence could be weak: if a state has no real stake in an issue, and its ally's actions will not affect reliability, then the state may not care how its ally behaves.⁴¹

Importantly, this concept of interdependence allows for the ally's actions to be important now, not just in future crises. Considering how interdependence operates simultaneously across discrete alliance commitments allows for the idea of reliability to be tested within a single crisis. Further, it is more representative of the real world: a state might not need to worry significantly about an adversary's resolve until the next crisis, but Brett Ashley Leeds finds that when “conditions change, [alliance] violation becomes more likely.”⁴² Thus, I expect states to be sensitive to variations in their ally's reliability.

However, because reliability is contingent upon interests—not character judgments—the effects of alliance interdependence might not be as severe as those predicted by other theories. If states do not assess character, then an ally's unreliability on one particular issue will not result in the ally being unreliable on all matters. A U.S. backdown and withdrawal in one region may not damage the United States’ reliability in another: it may, instead, improve it. Alternatively, the United States’ unwillingness to support a reckless ally may mean nothing for its determination to defend that same ally against an unprovoked attack.

The above analysis generates two hypotheses that can test whether alliance interdependence exists and whether it is influenced by reliability.

H1: A state will observe its ally's behavior in other alliances. If the behavior reveals divergent interests and thus raises entrapment or abandonment fears, the state will assess the ally to be unreliable.

H1 posits that convergent or divergent interests, not judgments of character, underpin alliance interdependence. If H1 is supported, then discrete alliance commitments are interdependent: the state is “using past behavior to predict or explain future behavior.”⁴³ But because the observing state is not assessing national character, not all observations will decisively revise reliability perceptions. Depending on the importance of the issue or interest, interdependence may be weak; but because the behavior still reveals new information about the ally's preference ordering, it will update the state's assessment of its ally's reliability. Interdependence will be strong, though, if the ally's behavior concerns an issue of significant importance to the observing state.⁴⁴

H2: If a state assesses its ally to be unreliable, it will act to mitigate the specific risk posed.

H2, in specifying the impact of interdependence, makes the concept of reliability satisfying. If the behavior observed in H1 is of significant importance, then interdependence will be strong, and as expected by H2, the state will attempt to mitigate the risk of allied unreliability.

H2 does not specify exactly how a state will respond, as this will depend on why the ally is unreliable. If the state fears abandonment, it might attempt to convince the ally that a more aggressive approach is wise or seek new allies that share its interests vis-à-vis the adversary. The state could boost its internal balancing by purchasing new equipment, building advanced weapons, stockpiling munitions and provisions, or introducing conscription. In severe situations, the fear of abandonment could prompt the development of nuclear weapons. If the state believes that its ally is growing more unreliable as time passes, it might seek to initiate conflict before the ally's reliability diminishes further. Alternatively, a weak state fearing abandonment may have no options beyond complaining to its ally or conciliating the adversary.

If the ally is unreliable because it poses risks of entrapment, the state could encourage the ally to adopt a more restrained policy by arguing that conflict is unwise. It might privately or publicly distance itself from the ally, signaling that it is unlikely to provide military support. If the allies share intelligence or military facilities, the state might restrict access to reduce the ally's likelihood of success and thus dissuade it from engaging in violence. Finally, the state could work with other nations to restrain the ally, launch diplomatic peace efforts, impose sanctions, or, in extremis, threaten to abrogate the alliance and defect to the opposing side in any resultant conflict.

In either scenario, allies are likely to discuss, and bargain over, shared interests. A state can attempt to convince its ally to order its preferences differently, for the sake of alliance unity, and the prioritization of interests might change as a result.

To test my theory against alternate explanations, I generate three hypotheses for deterrence theory (DT).

DT H1: A state, wanting to see its ally demonstrate loyalty, will observe the ally's behavior in other alliances.

DT H2a: If the ally demonstrates loyalty, the observing state will expect the ally to be loyal in the future.

DT H2b: If the ally demonstrates disloyalty, the observing state will expect the ally to be disloyal in the future.

Mercer's work, which is skeptical of reputation, is suitable for generating competing hypotheses about the independence of alliance commitments:

JM H1: A state, wanting to see its ally “behave properly” and demonstrate loyalty, will observe its ally's behavior in other alliances.⁴⁵

JM H2a: If the ally demonstrates loyalty, the observing state will not expect the ally to be loyal in the future.

JM H2b: If the ally demonstrates disloyalty, the observing state will expect the ally to be disloyal in the future.⁴⁶

Deterrence theorists encourage demonstrations of loyalty because disloyalty risks damaging other alliances. Mercer warns against fighting for reputation, because allies “will tend to assume that our interests and capabilities determine our resolve” and thus the United States will never gain a reputation for loyalty.⁴⁷ My theory treads a middle path, recommending that although the consequences of alliance interdependence should be considered, allies will not always want to observe loyalty (see table 1). In some cases, fighting to preserve allied assessments of reliability will be wise, as doing so will prevent undesired allied behavior. Sometimes, however, it will be wisest to back down to reassure allies fearing entrapment. There is no one-size-fits-all response.

The First Taiwan Strait Crisis, which began in September 1954 and concluded in April 1955, is a hard case for testing reliability, and should be an easy case for loyalty, for three reasons. First, the People's Republic of China (the PRC, or Communist China) was identified by the United States and its allies as an adversary, as was the PRC's ally, the Soviet Union. The defender, the Republic of China, on the island of Formosa (now Taiwan), had enjoyed informal ally status since the outbreak of the Korean War in June 1950, when President Harry Truman ordered the U.S. Navy's Seventh Fleet to defend the island.⁴⁸ If the United States was unwilling to stand by the ROC, then ideas of loyalty and interdependence should have prompted allied concern about future disloyalty.

Second, the crisis occurred when the United States was deeply worried about its reputation: in 1953 the National Security Council (NSC) requested an analysis of the “reported decline in U.S. prestige abroad,” and in November 1954, an assessment warned that because of the communist bloc's apparent victory at the Geneva Conference, U.S. “prestige … has suffered greatly.”⁴⁹ During the crisis, the “Never Again!” schema was a clear influence.⁵⁰ Days before the crisis, Secretary of State John Foster Dulles insisted that it was “highly important that we do not lose any more prestige in this area … [he was] willing to take a strong line militarily which might involve de facto belligerency for a certain period.”⁵¹ Accordingly, any behavior casting doubt on the importance of loyalty—either by allies or by U.S. decisionmakers—is a strong indication that other factors, such as reliability, have greater explanatory power.

Third, traditional arguments about loyalty and alliance interdependence carried the day for the majority of the crisis. The object of the dispute was a number of small offshore islands—predominantly Quemoy and Matsu— which are contiguous with the Communist-held mainland, more than 100 miles from the island of Taiwan. Although held by Nationalist forces in well-established defensive fortifications, the islands were easily within range of the PRC's artillery on the mainland. And despite never having been publicly guaranteed by the United States, they were regarded as sitting within the alliance halo. The United States could have justified abandonment by citing a strict definition of loyalty, but concern over the consequences of disloyalty dominated official thinking for seven months. Thus, any U.S. decision to reject these ideas about loyalty and interdependence would be doubly powerful: not only rejecting a general, preexisting belief that loyalty mattered, but also refuting the arguments that had previously determined U.S. policy in the crisis.

Beyond being a hard case, this crisis is also an extremely useful case for theory development. The United States was loyal to the ROC on the core issue of Formosa's security, but disloyal on the offshore island question, and this enables an examination of how allies interpreted contrary signals. Decisionmakers had to regularly reevaluate their reasoning and defend their decisions in the face of opposing arguments and new allied perspectives. Because the U.S.-ROC Mutual Defense Treaty was signed during the crisis, this case study also enables an exploration of how interdependence might vary across both informal and formal alliances. Because alliance formalization did not alter these alliance dynamics, this suggests that both informal and treaty alliances can be interdependent. Finally, the crisis was—from its opening salvos—one in which the real issues at stake were prestige and reputation.

Below, I use process tracing to test my idea of reliability against the alternative explanations. Using extensive archival evidence, I show how allies perceived the United States’ actions and sought to mitigate the risks posed by its unreliability.⁵² A single case study cannot generate generalizable findings, but it does strongly suggest that because alliance interdependence is not governed by loyalty, the consequences of disloyalty are easily misunderstood. After providing relevant background, I explain the chronology of the case study with reference to the views and reactions of Australia, Canada, New Zealand, and the United Kingdom.⁵³ I conclude the case study by examining the reactions of three other allies: Japan, the Philippines, and South Korea.

In 1949, the Truman administration attempted to absolve itself of responsibility for the ROC: an August White Paper distanced the United States from the Nationalists and their leader, Generalissimo Chiang Kai-shek.⁵⁴ In January 1950, President Truman confirmed that the United States would not use military force to restore Nationalist rule over mainland China. When Secretary of State Dean Acheson publicly described the United States’ defensive perimeter that same month, the Chinese Nationalists—holding the island of Formosa and the nearby Pescadore Islands—did not receive a security commitment. But when North Korea initiated the Korean War on June 25, 1950, this perimeter was quickly revised. The Seventh Fleet was dispatched to the Taiwan Strait: it was to deter communist attacks against Formosa, while also deterring any Nationalist effort to attack the mainland and widen hostilities.

By 1953, President Eisenhower had become frustrated by the lack of progress toward an armistice in Korea. On February 2, he announced that the Seventh Fleet would no longer “shield Communist China,” as the United States had no “obligation to protect a nation fighting us in Korea.”⁵⁵ U.S. policy, as defined in NSC 146/2, was to defend Formosa and the Pescadore Islands “even at grave risk of general war.” This document did not commit the United States to defending the offshore islands, but brought them within the alliance halo by instructing U.S. agencies to “encourage and assist the Chinese National Government to defend” them.⁵⁶

Prior to the crisis, decisionmakers in Washington recognized that these policies could damage other U.S. alliance relationships. In November 1953, NSC 166/1 cautioned, “The Free World will not act as a unit toward Communist China … the United States can avoid the most dangerously divisive potentials of the Chinese Communist issue, by refraining from excessive pressure on its friends to follow American policies.”⁵⁷ In 1954, the potential for allied disunity grew, as the United States proposed intervention in Indochina. A State Department official warned that Western unity could be threatened by “the fear that the U.S., in an excess of anti-communist zeal, may launch a military crusade against Communism and bring on World War III … [and] a fear that the U.S. may wash its hands of other nations and leave them to their fate … what other peoples are afraid of are the extremes of conduct to which we might resort.”⁵⁸ Dulles acknowledged that maintaining alliance unity would be difficult, and cautioned the NSC that it must “avoid getting the United States into a war … where world public opinion would be wholly against the United States, because that … was the kind of war you lose.”⁵⁹ Against this backdrop, on September 3, 1954, the PRC launched an artillery attack against the offshore island of Quemoy.⁶⁰

The purpose of the attack on Quemoy was not perfectly clear: although Communist China had earlier announced its intent to “liberate Formosa,” the opening salvo targeted only Quemoy.⁶¹ In the United States, the Joint Chiefs of Staff (JCS) disagreed about the military value of this island: the “majority opinion” argued it was “important but not essential to the defense of Formosa.” The more vigorous debate concerned the symbolic value of the offshore islands. The JCS worried about the psychological impact “on the Chinese Nationalist troops and other Asiatic countries inclined to support U.S. policy” if Quemoy were lost.⁶²

Adopting a domino theory lens, Dulles wrote that the “loss of Quemoy would … lead to mounting Communist action against deteriorating anti-Communist morale … this would be beginning of chain of events which could gravely jeopardize entire off-shore position.”⁶³ Dulles was especially reluctant to back down given his belief that the West had just done so in Indochina. According to Walter Robertson, the assistant secretary of state for East

Asia and the Pacific, Dulles “felt that if Quemoy can be held much of the Communist prestige stemming from Dien Bien Phu will be canceled.” Thus, the “U.S. should help hold Quemoy, even if its defense could not be substantially related to [the] defense of Formosa.”⁶⁴ In other words, Dulles advocated defending Quemoy purely to demonstrate loyalty and resolve.

Eisenhower worried about U.S. prestige, but was skeptical about fighting for reputation. Defending Quemoy would mean “our prestige is at stake … once we get tied up in any one of these things our prestige is so completely involved.”⁶⁵ The JCS and the State Department advocated defending the islands, but Secretary of Defense Charles Wilson expressed doubts and noted that strikes against the Chinese mainland would be an act of war, requiring congressional authorization. Vice President Richard Nixon suggested that the United States should help the Nationalists evacuate the islands, but Adm. Arthur Radford, chairman of the JCS, thought this would cause a collapse of morale on Formosa and its subsequent loss to Chinese Communist subversion.⁶⁶

When the NSC met again on September 12, Eisenhower decided that “Quemoy was not really important except psychologically.” He was “against making too many promises to hold areas around the world and then having to stay there to defend them … If we get our prestige involved anywhere then we can't get out.” By contrast, Dulles's comments highlight the different policy prescriptions the concepts of deterrence theory and alliance reliability might recommend. The secretary of state argued that an ROC retreat “would have disastrous consequences in Korea, Japan, Formosa and the Philippines,” but that if the United States was loyal to the ROC and defended the offshore islands, this “would involve us in war with Communist China. Outside of [South Korean President] Rhee and Chiang, the rest of the world would condemn us.” Equivocating, Dulles lamented that an “overwhelming case can be made on either side,” and recommended that a cease-fire be sought in the United Nations (UN) Security Council.⁶⁷

The United Kingdom, as the United States’ closest ally, quickly requested consultations. Dulles explained to British Foreign Secretary Anthony Eden that an “all out” Communist assault might capture Quemoy “unless [the] A-bomb [was] used tactically in [a] last resort.” The British echoed Nixon's suggestion that the Nationalists evacuate the islands. Although Dulles thought that this outcome was desirable “eventually,” he considered it “not practical now.” Dulles explained to Eden a concern that would dominate U.S. considerations throughout the crisis: disloyalty over Quemoy—such as a coerced evacuation—would cause Nationalist morale to plummet, the Chinese Nationalist regime to collapse, and Formosa to pass to Communist control.⁶⁸ Eisenhower had described Quemoy as “not really important except psychologically,” but this importance was so severe that U.S. decisionmakers now contemplated using nuclear weapons to defend the island.⁶⁹ Alarmed by this response, the United Kingdom encouraged Dulles's suggestion of a cease-fire resolution. Working with New Zealand, which held a nonpermanent seat on the UN Security Council, the three allies developed “Operation Oracle,” a plan for New Zealand to propose a Security Council cease-fire resolution. The allies expected this resolution to prompt a Soviet veto, but even this outcome would have the positive effect of highlighting the true aggressor in the crisis.

As work commenced on Oracle, Dulles reconsidered the ROC's long-standing desire for a bilateral alliance. The United States had previously refused this request, despite repeated efforts, given its fear of entrapment and its desire for Nationalist China to occasionally conduct small-scale attacks against the mainland. Dulles noted that the PRC's attacks against Quemoy meant that any alliance treaty would require careful drafting. Although President Eisenhower “was not ready to use the armed forces of the United States for the defense of these islands,” no treaty text could reveal this decision, because it would “green-light” a PRC invasion.⁷⁰

As part of Operation Oracle, U.S. officials briefed both the United Kingdom and New Zealand on the proposed alliance. When the British asked about the treaty area, a State Department official said that the islands would not be explicitly included but that the United States was “working on language which would cover that point.” Dulles privately emphasized that the ROC would have to promise not to initiate hostilities without U.S. approval, but the United Kingdom and New Zealand remained concerned about escalation. These two states decided that they were unwilling to advance Operation Oracle until Beijing's reaction to U.S.-ROC alliance negotiations could be observed. Officials in New Zealand were “anxious to know the terms of the proposed U.S. announcement concerning its treaty negotiations,” and encouraged U.S. officials to “emphasize as much as possible the defensive nature of the proposed treaty.” Dulles assessed that the United Kingdom would not commit to initiating Oracle “unless and until it felt that our [U.S.-ROC alliance] treaty undertaking would be compatible” with the cease-fire resolution plan.⁷¹

The State Department had earlier recognized the “desirability of keeping the Communists guessing” about the offshore islands, and the mutual defense treaty between the United States and the ROC, signed on December 2, 1954, was carefully drafted with this in mind.⁷² The treaty area was limited to Formosa and the Pescadores, but the activating clause referred to an attack “directed against” the territories of either party. Although Dulles argued that the status of the offshore islands was “neither promoted … nor … demoted by the treaty,” this wording allowed the United States to interpret an attack against an offshore island as the start of a military campaign directed against Formosa.⁷³ Assistant Secretary of State Robertson thought that this phrasing would “keep the Communists guessing as to what U.S. intentions are.”⁷⁴ While this commitment was made ambiguous, another was kept secret: the ROC agreed that it would not attack Communist China without Washington's approval.⁷⁵

The problem with the United States’ ambiguous stance was that it also kept its allies guessing. Dulles, Robertson, and Radford were all well-known China hawks. Earlier in 1954, the British thought that the JCS chairman “wanted to use [Indochina] as a means of launching what Eden said was ‘Radford's war against China.'”⁷⁶ Robertson, described by one historian as “rabidly anti-Communist,” was regarded by the British as a “loathsome figure.”⁷⁷ In August 1954, a State Department official wrote that “one of the chief reasons why so many countries are concerned lest U.S. policy lead to general war is because … [we support] the claim of the Chinese Nationalists to constitute the true Government of China—a claim that cannot be made good except as the result of all-out war between the U.S. and Communist China.” For allies, “the central political issue … was the worry that Washington was ‘too quick on the trigger.'”⁷⁸ Although U.S. officials insisted that they did not intend to defend the offshore islands, Dulles had candidly admitted to Eden that the United States was “keeping [the Communists] … guessing partly because we're guessing ourselves.”⁷⁹ So even as hawkish U.S. officials professed their desire for a cease-fire, allies had cause for concern.

Deterrence theory would expect allies to desire loyalty (DT H1), for loyal behavior to create expectations of future loyalty (DT H2a), and for disloyal behavior to create expectations of future disloyalty (DT H2b). In contrast to DT H1's expectations, both the United Kingdom and New Zealand worried about a surfeit of U.S. loyalty toward the ROC. Supporting DT H2a, they worried that U.S. loyalty to Nationalist China would continue and result in conflict. Finally, if these countries believed that disloyal treatment of the ROC suggested future U.S. disloyalty, they would not have encouraged the United States to organize a Nationalist withdrawal. The intensification of entrapment concerns in London and Wellington, and their efforts to restrain the United States, falsify DT H1, support DT H2a, and cast doubt on the logic underpinning DT H2b.

Mercer would expect allies to desire U.S. loyalty to the ROC (JM H1), that loyal behavior would not cause allies to expect future loyalty (JM H2a), and that disloyal behavior would cause allies to expect future disloyalty (JM H2b). Contrary to JM H1, the United Kingdom and New Zealand were utterly unconcerned about possible disloyalty to the ROC. Instead, they worried that U.S. loyalty to the ROC suggested that the United States would be similarly loyal in the future, thus increasing the risk of war. This evidence falsifies JM H2a, which expects that a demonstration of loyalty will not cause observing states to expect similar loyalty in the future. Finally, allied officials were “anxious to restrain Washington” and so encouraged disloyalty to the ROC.⁸⁰ If JM H2b were true, and these states believed present disloyal behavior to be suggestive of future disloyalty, they would not have encouraged an evacuation. The events examined so far falsify JM H1 and JM H2a, but because the United States had not demonstrated clear and unambiguous disloyalty toward the ROC, I do not yet describe DT H2b and JM H2b as falsified.

In contrast, these events do support the alliance reliability concept: as expected by H1, allies observed the United States’ behavior toward the ROC and updated their beliefs about U.S. reliability. Allies expressed concern about the United States’ determination to assist the ROC and advocated a withdrawal from the offshore islands. Because of this divergence in interests, allies worried about escalation and believed that U.S. reliability had declined. In response— and as expected by H2—these states worked to mitigate the risk of entrapment. For now, their efforts focused on private consultations with officials in Washington. Through these actions, they had commenced a monthslong effort to restrain the United States and to persuade its leaders that alliance cohesion required disloyalty to the ROC.

On January 10, 1955, the PRC launched an air attack against Nationalist forces on the Tachen Islands, located 200 miles north of Taiwan. Days later, it attacked and occupied the Yikiangshan Islands, only eight miles from the Tachens. Secretary of State Dulles worried that U.S. inaction was “having a bad effect on our prestige in the area, since it was in many quarters assumed that we would defend the islands, and our failure to do so indicated that we were running away.” This remark is especially revealing: Dulles believed that the United States would be judged by allied expectations of its loyalty toward the ROC, rather than by whether the United States lived up to its formal security commitment. Eisenhower decided that the United States should offer to evacuate ROC forces from the Tachens, but then publicly commit to defending Quemoy.⁸¹ Dulles promptly briefed the ROC's ambassador on this idea, while also hoping that a cease-fire resolution (Operation Oracle) would be moved in the UN Security Council.⁸²

British officials, however, criticized this plan, believing that it would “encourage the Nationalists to hang on to the islands.” Deftly, they recalled Dulles's earlier assessment—that holding Quemoy required nuclear weapons—and asked “whether Quemoy was sufficiently vital to risk such wide-reaching developments.” Dulles replied that Quemoy must be guaranteed, because “if the Tachens are evacuated and no other move or explanation given, the impression will be that of a collapse in position. The consequences … in Japan, Korea, the Philippines and very possibly throughout Southeast Asia would be extremely serious. The interpretation which would gain currency was that the U.S. was vague until its interests were attacked and then did nothing.” Robertson agreed, arguing that U.S. officials were considering “morale throughout all of free Asia.” Finally, Dulles attempted to allay British concerns, promising that his earlier reference to nuclear weapons “related only to the most extreme hypothesis of … a human wave” attack, which was a “remote possibility.”⁸³

Dulles failed to convince the allies, however, and returned to the NSC empty-handed. The best possible compromise was that the United Kingdom and New Zealand might support the U.S. plan if there were no public security commitment to any offshore island. Concerned that the U.S.-ROC treaty text might prove insufficient from a domestic politics standpoint, the NSC agreed to seek a congressional resolution empowering President Eisenhower to defend “Formosa and the Pescadores against armed attack,” and that this would “include the securing and protection of such related positions now in friendly hands.”⁸⁴ As Robert Accinelli notes, a lack of allied support influenced U.S. policy: the undertaking to defend Quemoy, regarded by allies as “ill advised and dangerous,” was kept “secret as a concession to the British.”⁸⁵

When Chiang Kai-shek learned that the United States was reneging on its pledge to publicly guarantee Quemoy, he refused to evacuate the Tachens. The United States’ response was that its promise to defend Quemoy and Matsu—in exchange for the evacuation of the Tachens—must remain private and might be disavowed if it was leaked.⁸⁶ With this confirmed, the United Kingdom and New Zealand initiated Operation Oracle, called for a cease-fire, and invited a PRC representative to attend the UN Security Council. When Beijing denounced the resolution, these tensions prompted other allies to increase their own efforts. In early February 1955, the Australian prime minister stated that “opinion would be much opposed to accepting a risk of war over the ‘off-shore’ islands.”⁸⁷ A State Department official wrote to Dulles, assessing that Commonwealth allies could support the United States “if we … have the Chinese Nationals withdraw from all the off-shore islands and make our stand on Formosa.”⁸⁸ Significantly, as the crisis persisted, allied criticism would shift from private words of caution to more public discussion of the risks involved.

Reluctantly accepting the U.S. policy reversal, the Nationalists requested assistance to evacuate the Tachens. The withdrawal was completed by mid-February, but the broader dilemma remained: What would the United States do if the PRC attempted to capture Quemoy and Matsu? Eisenhower wrote Prime Minister Winston Churchill, insisting that the psychological stakes were critical: “Korea, Japan, the Philippines, and, of course, Formosa itself, are watching nervously to see what we do next. I fear that, if we … compel Chiang to make further retreats, the conclusion of these Asian peoples will be that they had better plan to make the best terms they can with the Communists.”⁸⁹ Dulles tried a similar approach with Foreign Secretary Eden, telling him that “we are in a battle for Taiwan,” Dulles invoked the need to demonstrate loyalty: “Further retreat would have grave effect on Taiwan and in Asia … Further retreat could swing Asia. Trends in Japan are already disturbing. Further retreat or loss of Formosa would convince Japan communism wave of future. Consequent effect on … other parts of Asia obvious.”⁹⁰ The British leaders, however, were unconvinced and again recommended withdrawal from the islands.

Allied reactions now cast serious doubt on the validity of deterrence theory, which expects allies to desire loyal behavior (DT H1), for loyal behavior to create expectations of future loyalty (DT H2a), and for disloyal behavior to create expectations of future disloyalty (DT H2b). Events examined so far support DT H2a, but falsify DT H1 and DT H2b. When the United States displayed loyalty to the ROC through public signals such as the Formosa Resolution, and its promise to publicly guarantee the security of Quemoy, other allies worried that this increased the likelihood of future loyalty and thus raised the risk of conflict: this supports DT H2a. These allies were pleased, however, when the United States was disloyal and refused to publicly commit to a defense of Quemoy. If DT H2b were true, then allies would not have welcomed this, because it would have suggested a character trait of disloyalty. Finally, allied encouragement of disloyalty not only falsifies DT H1, but also suggests that allies were not judging character. If allies believed that disloyalty to the ROC was proof of the United States’ disloyal character, they would not have encouraged such behavior.

Mercer would expect allies to desire U.S. loyalty to the ROC (JM H1), that loyal behavior would not generate allied expectations of future loyalty (JM H2a), and that disloyal behavior would generate allied expectations of future disloyalty (JM H2b). The reactions of the United Kingdom, New Zealand, and Australia clearly falsify JM H1: these states encouraged U.S. disloyalty to the ROC. They observed instances of the United States’ loyalty to Nationalist China and, as a result, worried about the likelihood of future loyalty, thus falsifying JM H2a. In contrast to the expectations of JM H2b, there is no indication that these states expected U.S. disloyalty toward the ROC to be suggestive of future disloyalty, and thus JM H2b is also falsified.

In contrast, allied responses clearly support the idea of reliability. As expected by H1, U.S. behavior revealed divergent interests and caused allies to grow more concerned about the possibility of escalation and entrapment. In response, these countries encouraged U.S. disloyalty toward the ROC. Like-minded allies presented a united message: a Nationalist withdrawal from the offshore islands would increase their support for U.S. policy. As a result of this allied lobbying, the United States reneged on its promise to the ROC and did not make an explicit public commitment regarding Quemoy. Through such intra-alliance bargaining, states worked to mitigate U.S. unreliability and the risk of entrapment, and this evidence supports H2.

In a bold display of brinkmanship, on March 8, 1955, Secretary of State Dulles stated that the United States regarded nuclear weapons as “interchangeable with … conventional weapons.” Within days, similar sentiments were expressed by President Eisenhower and Vice President Nixon, but these threats found no allied support.⁹¹ Australia suggested that the ROC might be persuaded to withdraw from the offshore islands if, in exchange, the U.S.-ROC alliance were expanded to include other nations such as Australia, New Zealand, and the United Kingdom.⁹² On March 8, in the House of Commons, Foreign Secretary Eden “openly advocated a Nationalist withdrawal from Quemoy and Matsu.”⁹³ He later privately explained that “not one percent of British people” would support a conflict over the offshore islands, and he could not “increase that percentage no matter how hard I tried.”⁹⁴ The Canadians cautioned Dulles that if nuclear weapons were used against the PRC, then the Americans “would be [fighting] on their own as far as Canada was concerned.”⁹⁵ Fearing entrapment, allies dramatically warned—both in private and in public—that they would not fight for the offshore islands.

In late March, Eisenhower was still standing firm. He wrote a friend, “If you became convinced that the capture of [Quemoy and Matsu] … would inevitably result in the later loss of Formosa to the free world, what would you do?”⁹⁶ Even though some of the domino states were, in their willingness to cede the offshore islands, rejecting the domino theory, U.S. officials continued to fret.⁹⁷ But in late March, a new survey of Nationalist morale challenged Eisenhower's position with the conclusion that “subversion [on Formosa] is well under control … Chinese-American cooperation continues to be satisfactory.”⁹⁸ If Nationalist morale could withstand the loss of Quemoy and Matsu, then the ROC government would not collapse, and U.S. disloyalty on this issue would not damage other important alliances.

In early April, Eisenhower's views evolved and he took a more skepticaltone.⁹⁹ After noting the alliance damage that would flow from a defense of the offshore islands, Eisenhower wrote that withdrawal “might have equally disadvantageous results” and that “further retreat … could result, it is alleged, in the disintegration of all Asian opposition” to communism. In this letter, Eisenhower describes falling dominoes as a possibility, rather than a certainty.¹⁰⁰ This judgment enabled a policy shift: Eisenhower decided the United States would not defend the offshore islands, and so he encouraged a Nationalist withdrawal. In return, the United States would station a Marine division on Formosa, try to expand the U.S.-ROC alliance to include other states, and blockade a portion of Communist China's coastline. Here, Eisenhower decided that maintaining reliable alliances was more important than preserving a reputation for loyalty.

Robertson and Radford were dispatched to Taipei to convince Chiang Kai-shek to accept this arrangement, but he was immovable. He insisted that “soldiers must choose proper places to die. Chinese soldiers consider Quemoy—Matsu are proper places for them.”¹⁰¹ Ultimately, the United States escaped its dilemma “not through the ill-conceived evacuation-blockade scheme but an unexpected offer from the Chinese Communists.”¹⁰² On April 24, PRC Premier Zhou Enlai stated that he was willing to discuss security tensions directly with the United States. Although Dulles was initially skeptical, allies quickly encouraged direct negotiations, and when the United States agreed to direct talks with the PRC, “Washington's openhanded response … brought an end to the Taiwan Strait crisis.”¹⁰³

To explain the main dynamics of the First Taiwan Strait Crisis, I have focused so far on the perceptions and actions of the most influential allies. Although less prominent, other allies also reacted as the alliance reliability concept would expect.

Japan was petrified about a possible war with Communist China. Secretary of State Dulles repeatedly claimed that Japan worried about retreat, but the U.S. embassy in Tokyo reported Japan's “desire for peace at almost any price,” noting that “such lesser matters as … which side loses ‘face’ are relatively unimportant.”¹⁰⁴ The U.S. ambassador to Japan, John Allison, wrote that if the United States defended the offshore islands, most Japanese would view “the U.S., rather than Communist China … [as] the principal ‘war monger.'” Japan would approve a defense of Formosa itself, if necessary, but was especially worried about the entrapment risks posed by the offshore islands.¹⁰⁵

Under the bilateral alliance signed in 1951, the United States had the legal right to use bases in Japan—even without permission from the Japanese government—for any purpose the United States deemed relevant to Far Eastern security. Thus, Japan's entrapment risks were far more severe than those faced by other countries. Ambassador Allison assessed that if the United States used these bases, then Japan would “take more positive steps to ensure Japanese non-involvement,” such as “failure to act officially against … [labor] strikes at US bases or … [a] request not to use Japanese bases for the staging of attacks on the China mainland.” Ominously, Allison warned that if the United States used nuclear weapons, this could cause a “sharp swing of Japan into the ranks of the neutral nations [and] … might well jeopardize our whole position” in Japan, as Tokyo would request the withdrawal of all U.S. forces.¹⁰⁶ By contrast, Allison's remarkable assessment was that if the offshore islands fell because the United States stood aside, this disloyalty was “likely to be greeted with relief, even praised,” in Japan.¹⁰⁷

Given its unique and unequal postwar position, Japan could not quickly mitigate the entrapment risks posed by the United States’ excessive loyalty to the ROC. It did, however, use public messaging, as well as private diplomacy, to make its position clear. Shortly after the crisis ended, Japan quickly moved to renegotiate its alliance with the United States. The final result—the 1960 Treaty of Mutual Cooperation and Security between the United States and Japan—sharply reduced U.S. military freedoms and mitigated the risk of Japan's main islands being dragged into a Taiwan Strait conflict. Although domestic politics was a significant motivation for treaty revision, recent research has also noted the influence of Japan's entrapment fears.¹⁰⁸

The Philippines’ reaction was affected by geographical proximity to Formosa, misunderstanding, and excessive confidence in U.S. military capabilities. Foreign Secretary Carlos Garcia, reacting to the PRC's attack against the Tachen Islands, “expressed alarm for [the] safety [of the] Philippines as [a] result [of the] Tachen incidents.” He briefed the press that Manila was “‘closely watching’ Tachen development,” and that the “entire free world's faith in America will hinge on your ability to cope with [the] situation.”¹⁰⁹

Despite this apprehension, there were no “serious reactions” to the Tachens withdrawal.¹¹⁰ In early April 1955, Felino Neri, a foreign affairs adviser to Philippines President Ramon Magsaysay, implored U.S. diplomats to provide a clearer explanation of United States policy and encouraged them to “understand [the] Asian mind. To us a retreat on Quemoy and Matsu means a retreat in all of Asia.”¹¹¹ The State Department assessed that “sentiment in the Philippines is very strongly against … abandonment of Quemoy and Matsu. The islands are generally thought to be important to the defense of Taiwan. Loss of the islands … would cause serious concern … that the United States was unwilling, unable, or both, to fulfill its Pacific commitments.”¹¹² Later in April, Neri complained that “many Filipinos fear U.S. determination is wavering, especially under pressure [from] British, Canadian allies, and that failure [to] defend Quemoy, Matsu would lead inevitably to withdrawal from Formosa and Philippines.” The U.S. embassy in Manila assessed that these fears, though genuine, rested on a misunderstanding: beyond overestimating the importance of Quemoy and Matsu, “most Filipinos … do not include serious considerations of world war possibilities, but rather take for granted that … Formosan issue can be limited to local action.”¹¹³ Given these beliefs, it is unsurprising that Manila desired a defense of the offshore islands. Other allies—with different understandings of the military situation and risks—were more worried about escalation.

Despite these fears, the Philippines was “highly elated” when the PRC hinted at a cease-fire.¹¹⁴ The available evidence suggests that the Philippines carefully monitored the U.S.-ROC alliance, and that these observations caused it to doubt U.S. reliability: the Philippines feared that abandonment of Nationalist forces on Quemoy and Matsu might be replicated in the future. Perhaps because of an excessive faith in U.S. military power and an expectation of only limited hostilities, Filipino leaders encouraged a defense of the offshore islands.

Among United States allies, the Republic of Korea was the unambiguous outlier: Seoul wanted conflict, as this would provide an opportunity to restart the Korean War. One U.S. diplomat thought that South Korea's foreign minister, Pyun Yong Tae, held the “basic premise that World War III is inevitable and believes sooner the better.”¹¹⁵ At a press conference, the Korean ambassador to the ROC said he was “sure” that in the event of an attack against Formosa, the United States would “allow [South] Korea to start its ‘long withheld military drive into North'” Korea.¹¹⁶ Such data suggest that Seoul advocated U.S. loyalty to the ROC not because it would confirm national character, but because it could result in wider hostilities. In September 1954, Pyun sent a message to Dulles: “In this part of world Quemoy can be symbol, loss of which … would have serious repercussions in Asia.”¹¹⁷ Although these words suggest that South Korea viewed the crisis as a test of U.S. reliability, the United States was ultimately unmoved by South Korea's pleas. When Washington opened direct talks with Beijing, the South Korean press claimed that Eisenhower was “risking alienation of America's only real friends in the Far East.”¹¹⁸

Deterrence theory would expect states to desire allied loyalty and observe their ally's behavior to make judgments about its loyalty (DT H1), for loyal behavior to create expectations of future loyalty (DT H2a), and for disloyal behavior to create expectations of future disloyalty (DT H2b). For the five allies fearing entrapment—Australia, Canada, Japan, New Zealand, and the United Kingdom—DT H2a is supported, whereas DT H1 and DT H2b are falsified.

The two allies desiring a stronger U.S. response—South Korea and the Philippines—partially fulfill the expectations of deterrence theory. As predicted by DT H1, these allies wanted to defend the offshore islands. Examples of United States loyalty to the ROC, however—such as the formalization of the U.S.-ROC alliance and the issuing of nuclear threats to the PRC—did not create expectations of future loyalty, so this falsifies DT H2a. Instead of being reassured, these allies thought that U.S. disloyalty in the present—illustrated by the United States’ refusal to actually defend the offshore islands—was suggestive of future disloyalty as well, and these data support DT H2b.

Mercer's reputation-skeptic approach expects that states want to observe allied loyalty (JM H1). If the ally demonstrates loyal behavior, the observing state will not expect the ally to be loyal in the future (JM H2a), but if the ally demonstrates disloyal behavior, the observing state will expect the ally to be disloyal in the future (JM H2b). As explained earlier, Mercer's hypotheses are falsified by the reactions of the five allies fearing entrapment. These allies encouraged U.S. disloyalty and showed no concern that this behavior might suggest future disloyalty. By contrast, the reactions of South Korea and the Philippines strongly support Mercer's theory. Both allies desired U.S. loyalty to the ROC (JM H1), appeared to take no comfort in the loyal behavior that the United States did demonstrate (JM H2a), and worried that Washington's unwillingness to defend the offshore islands was suggestive of future disloyalty as well (JM H2b).

The United States was neither perfectly loyal nor consistently disloyal to Taiwan: it was loyal in formalizing the U.S.-ROC alliance, in signaling a firm intent to defend Formosa, and in rattling the nuclear saber. But the United States was also disloyal: it did not meet the ROC's hope for counterattacks against the mainland; it reneged on its promise to publicly announce a security guarantee for Quemoy; and it ultimately attempted to cajole the Nationalists into withdrawing from the offshore islands. It is difficult, then, to simply describe the overall conduct of the United States as either loyal or disloyal: it was, in fact, both. Of greater interest is how U.S. allies interpreted this behavior, and what they did in response. The reactions of South Korea and the Philippines suggest that if states do want their allies to display innate and pure loyalty, then Mercer's theory has greater explanatory power than deterrence theory. These allies took no comfort in displays of U.S. loyalty and, instead, lamented that the United States was not being loyal enough. The other five allies, by contrast, thought that the United States was being too loyal to the ROC and worried about consequent risks of entrapment.

These differing reactions support the central claim of the alliance reliability concept: allies did not want a display of unalloyed loyalty to the ROC because it would reveal the national character of the United States. Rather, they wanted U.S. preferences to be identical to their own: they wanted reliability. Five allies desired a reduction of tensions and a Nationalist withdrawal from the offshore islands, but South Korea and the Philippines were disappointed by the United States’ disloyal behavior toward the ROC. These reactions all support H1: a state observes its ally's behavior in other alliances, and this behavior affects the state's assessment of the ally's reliability. If the United States behaved contrary to the ally's desires, because it was either too timid or too belligerent, it was perceived as unreliable.

The varied preferences of these allies presented them with different dilemmas. As expected by H2, all allies tried to mitigate the unique risk posed by U.S. unreliability. Both South Korea and the Philippines worried about U.S. timidity, but as small allies extremely reliant on their alliances, they had little capacity to influence the United States. As Miller notes, “If a state has only one potential ally, that ally's reputation will be a minor factor” in alliance politics.¹¹⁹ Their actions suggest, however, that when faced with similar indications of declining reliability, stronger allies could react in a different fashion.

Japan could not quickly mitigate U.S. unreliability, but once the crisis subsided, it moved to renegotiate its alliance and reduce the risk of entrapment. For other allies, Operation Oracle was “a useful instrument to monitor and moderate U.S. policymaking.”¹²⁰ New Zealand's high commissioner in London assessed that for the United Kingdom, “a major reason” for participating in Operation Oracle “was to place some kind of restraint on the United States.”¹²¹ New Zealand's ambassador in Washington said that Eden “always took the line that the Russians … would restrain the Chinese, and the British … would restrain the Americans.”¹²² These diplomatic efforts escalated as the United States grew more belligerent and unreliable: private discussions progressed to public warnings and, eventually, to distancing and outright disassociation. These actions support H2: when confronted with an ally's unreliability, states move to mitigate this risk.

Throughout the First Taiwan Strait Crisis, the belief that alliance interdependence was underpinned by loyalty proved remarkably resilient. Even after diplomatic reporting confirmed that several allies were opposed to a defense of Quemoy and Matsu, U.S. leaders continued to talk and act as if abandonment of the offshore islands would tear the Asian alliance system apart. President Eisenhower's eventual decision to back down was, to some degree, prompted by new thinking about the relationship between Quemoy and Matsu and the state of morale on Formosa. Eisenhower, however, also knew that U.S. loyalty to the ROC was damaging other, more important alliance relationships. Alliance commitments were interdependent, but not in the way expected by deterrence theory: several allies urged the United States to be disloyal toward the ROC and coerce the Nationalists into withdrawing from the offshore islands, because this disloyalty would reduce the risk of an unwanted war. Although the perceived need to demonstrate loyalty was an immense influence throughout the crisis, the need to demonstrate reliability was ultimately of greater importance.

The prolonged duration of the crisis was key in enabling U.S. decisionmakers to revise their assessments of what was at stake. It took months to correct the mistaken belief that U.S. national character was on trial in the court of allied opinion. If the PRC had attempted an invasion of Quemoy or Matsu, then President Eisenhower and Secretary of State Dulles—believing that the alliance system would be destroyed by a display of disloyalty—would have probably intervened. Several U.S. allies would likely have regarded such intervention as proof that the United States was intent on a final showdown with the PRC. Some may have accepted their entrapment into this conflict, but others might have ended their alliance with the United States.

The events of the First Taiwan Strait Crisis illustrate the need to delineate between innate loyalty and observed reliability. Reliability is a broader and more useful concept that subsumes the idea of loyalty and reduces its unwarranted prominence. U.S. conduct was not judged against an objective standard of loyalty: allies desiring a confrontational posture, such as South Korea and the Philippines, invoked Quemoy and Matsu as symbols that required defending, and encouraged U.S. loyalty to the ROC. Most allies, however, were more concerned about the possibility of war, and so discouraged a U.S. defense of strategically unimportant territory. These allies did not mind that this would involve U.S. disloyalty to the ROC, but instead cared about whether Washington's policy goals were convergent with their own.

In this case study, several allies cooperated to restrain the United States and mitigate the risk of entrapment, while weaker allies fearing abandonment had little option but to complain about U.S. timidity. In other circumstances, however, allies fearing abandonment could respond differently. For example, in May 1950 Japan dispatched a secret delegation to Washington. Citing the concern prompted by the 1949 withdrawal of United States forces from the Korean Peninsula, and vacillating U.S. policy toward Nationalist China, Japan expressed its desire for U.S. forces to stay in Japan, even once the postwar occupation ended.¹²³ Similar fears of abandonment were sparked in the 1970s, when President Jimmy Carter planned troop reductions in South Korea. In response, Tokyo initiated a new period of military and economic cooperation with Seoul. As Victor Cha writes, “The most single decisive factor in this upswing in security contacts was anxiety over the Carter plan.”¹²⁴

As explained, Mercer's argument about reputation is contingent upon what constitutes desired and undesired behavior. If states do not always desire allied loyalty, then this leaves open the possibility of interdependence being governed by assessments of interests and capability, but raises the question of when such interdependence might be observed. This article deployed a within-case study approach, but interdependence was also operative across iterative crises: allies were unsurprised by the United States’ initial response to the First Taiwan Strait Crisis, because they had just observed a similarly belligerent approach toward the PRC during discussions on Indochina. Although the United States ultimately backed down on Indochina in 1954, this did not lead its allies to expect a similar back down over the Taiwan Strait. This logic supports the argument made by Alex Weisiger and Keren Yarhi-Milo: by focusing on iterative security crises, previous research may have overlooked the possibility that during such events, “information gleaned from past actions … will have already been folded into the general assessment of interests.”¹²⁵ Studying prolonged security crises, where assessments of interests are regularly updated in response to new information, may provide fresh insight into not only how states assess allied reliability, but when and why these assessments change.

Although this article focuses on alliances, my findings raise new questions about the possible interdependence of threats. One assumption contested in this article—that allies always want to see their ally demonstrate innate loyalty—has an equivalent in deterrence theory. It is assumed that states always want their adversary to back down, but are there circumstances in which an adversary's decision to fight is welcomed?¹²⁶ A state may desire this if it saps public support, provides geography unfavorable to the adversary, or ties down an adversary's military force in an unimportant theater. China may not object to a display of U.S. resolve in the Middle East or Europe if it believes that this decreases the United States’ focus on Asia. This counterintuitive possibility requires further consideration and, if supported, would be further evidence to suggest that states assess national interests, rather than national character.

The policy implications of the alliance reliability concept are significant, especially given President Donald Trump's mercurial approach to alliance management. How should U.S. policymakers approach the possibility of alliance interdependence? First, they should realize that the national character of the United States is not on trial: allies will be looking for evidence of reliability, not loyalty. The idea of reliability is especially pertinent to current discussions about a U.S. withdrawal from the Middle East: allies may not invariably despair at withdrawal if it enables the United States to focus on other issues of greater allied concern. Second, U.S. policymakers should consider how specific actions might influence each ally, as different allies are likely to react in different ways. Officials will not need a crystal ball: this First Taiwan Strait Crisis case study suggests that allies will not hesitate to express their fears. Nor will allies rush for the door at the first sign of unreliability: in this case, their first instinct was to bargain with U.S. leaders and convince them that disloyalty to the ROC was actually in the best interests of the United States. The reliability concept shows that state interests are not totally exogenous to their alliance relationships: alliance unity can require hard bargaining over how states order their interests.¹²⁷ This bargaining is made more difficult if states in alliance are reluctant to consider situations where their interests diverge, even privately, for fear of upsetting alliance comity. Such discussions may be difficult, and frustrating for the United States if it expects unqualified support, but they provide an opportunity to understand what, in alliance interdependence terms, is at stake in a crisis. Finally, the United States should consider the desirability of likely allied reactions and factor these into decisionmaking. Fighting to preserve allied beliefs about U.S. reliability may be worthwhile if doing so prevents unwanted reactions such as dealignment, bandwagoning, or nuclear proliferation. On the other hand, it may be beneficial for allies to fear unreliability if this will almost certainly cause them to pursue policies—such as increased defense spending—desired by the United States.

Today, several flash points around the globe could create problems of alliance interdependence. The closest parallel to the events of 1954–55 is found in the Sino-Japanese dispute over the Senkaku/Diaoyu Islands, which are covered by the U.S.-Japan alliance. Another similar case is the recent reaffirmation, by Secretary of State Mike Pompeo, that the U.S.-Philippines alliance covers Filipino military assets in the South China Sea.¹²⁸ The instinct of many in the United States will be to regard any Chinese challenge in these areas as a test of national character. Instead, policymakers should consider their options and the likely allied reactions. Given the escalation risks present within the U.S.-China relationship, will other allies want to risk conflict—perhaps even nuclear war—over uninhabited territory of debatable strategic importance or the sinking of a single naval vessel?¹²⁹ Of course, arguments advocating disloyalty would have to be weighed against the likely effects in Japan, or the Philippines. Would Japan abrogate the alliance, conciliate China, or perhaps develop its own nuclear weapons? Or could it be reassured, thus avoiding undesired consequences? Would the Philippines react by bandwagoning with China? These questions will be even harder to answer during a security crisis, but neither deterrence theory nor reputation-skeptic perspectives would prompt decisionmakers to ask them. The concept of reliability and its effect on alliance interdependence do not solve such dilemmas, but they help to raise the right questions and more accurately identify the actual stakes involved.

A fourth Taiwan Strait crisis would probably be over territory of greater strategic significance: Taiwan itself. U.S. officials would probably not have the luxury of time to reconsider their beliefs about reputation, and decisions could be strongly influenced by the unexamined conviction that allies will desire a display of innate loyalty. Instead, decisionmakers should carefully investigate—before any crisis erupts—the possibility that allies may actually want the United States to adopt a more measured policy. If loyalty to Taiwan would create substantial risks of nuclear war, these allies may even prefer that the United States simply stand aside and allow Taiwan to fall under PRC control. Policymakers should carefully identify the allied interests that would be at stake, and such assessments should inform their considerations of what, in terms of alliance interdependence, would be at risk. They should also be sensitive to the possibility that national positions could shift: some states might initially encourage a firm stance, but if tensions escalate and nuclear risks emerge, then these states might change tack and belatedly encourage disloyalty.

This is not to blithely suggest that the circumstances of 1954–55 will repeat, but it is possible that there would again be a range of views among allies. Japan, fearful of the strategic implications of a Chinese victory, might encourage a firm U.S. response. In extreme scenarios, if U.S. reliability could not be assured and extreme forms of abandonment were regarded as likely, then Japan might develop its own nuclear capabilities. As Alexandre Debs and Nuno Monteiro have argued, if “the Trump administration question[s] U.S. commitments to its East Asian allies, Tokyo may be pushed to change its policy of nuclear forbearance.”¹³⁰ The events of 1954 and 1955 show that not all allies are of equal importance to the United States, and that the concerns of some allies may be prioritized while others are disregarded. The protection of the U.S.-Japan alliance may be so important that consequent damage to other alliances would be acceptable.

That said, the likely effect on other U.S. allies should not be ignored. In 1954–55, the concerns of South Korea and the Philippines could be safely discounted, because neither country had feasible realignment options. In any fourth Taiwan Strait crisis, allies desiring U.S. restraint could again decline to support the United States in a conflict. Given China's economic influence, and the prospect of a clash over Taiwan escalating into a nuclear war, there could be several such allies. The United States could easily find itself fighting alone, except (possibly) for Japan. It might end up reassuring Japan, but at the cost of dramatically damaging its other alliances in the region. Policymakers would need to consider alliance reliability carefully, for it is a double-edged sword: capable of cutting both ways against allies with different fears.

This article has advanced the debate about alliance interdependence by further developing the idea of reliability. Deterrence theory counsels policymakers to fight for reputation, because doing so strengthens alliances. Reputation skeptics such as Jonathan Mercer advise the United States to not worry about the views of allies, because the United States “cannot manipulate our allies’ image of us,” and “there probably will not be any future costs” resulting from U.S. behavior.¹³¹ These theories do not recognize, however, that alliance interdependence is underpinned by beliefs about reliability. In the First Taiwan Strait Crisis, excessive loyalty to Nationalist China caused other allies to regard the United States as unreliable. Had conflict broken out, then future costs would have been very likely: these other alliances may have been severely damaged or—in the case of Japan—perhaps even abrogated. Reputation-skeptic arguments are correct that a reputation for innate loyalty does not deserve the central importance afforded it by deterrence theory, but this does not mean that any prospect of alliance interdependence should be dismissed. Skillful alliance management requires decisionmakers to understand that the United States’ actions will not be perceived in a uniform manner by a united audience of allies. Although deterrence theory encouraged the conception of U.S. allies as a monolithic bloc—all desiring a display of loyalty—President Eisenhower realized that a firm stand could severely damage, or destroy, other important alliances. To manage future crises with similar adroitness and wisdom, U.S. decisionmakers will again need to distinguish between loyalty and reliability.

The author thanks the anonymous reviewers, David Capie, Timothy Crawford, Evelyn Goh, Michael Green, Yasuhiro Izumikawa, Yuen Foong Khong, Tongfi Kim, Amy King, Ronald Krebs, Gregory Miller, Evan Resnick, Jennifer Spindel, and Hugh White for their comments and suggestions. The author also thanks the staff at the National Archives and Records Administration, Princeton University's Seeley G. Mudd Library, and the Dwight D. Eisenhower Presidential Library. The author gratefully acknowledges the financial support provided through grants and scholarships from the Australian Government, the Australian National University, and the Australian-American Fulbright Commission.