| It doesn't stop|
at the water's edge
Deterrence is the art of convincing someone not to do something. In international relations, this is accomplished through the use of threats to inflict great harm in the event of a military attack. As such, it is more than simply being willing to fight if attacked: the term is nowadays used almost synonymously with the treat of nuclear retaliation, threatening massive destruction. As a broader strategy, deterrence is usually contrasted with appeasement. Traditionally, only threats of military action are truly considered deterrence; however, in less formal settings, it can refer to any threat. For example, the death penalty (and lesser punishments) are usually held up as deterrents against crime. An effective deterrent requires both a capability and credibility in order to be taken seriously.
For a threat to work, the one making the threat must be able to go through with the threat. For example, while Finland could threaten to nuke the United States if the US attacked Sweden, the US can feel confident that Finland is not actually capable of doing so. On the other hand, Finland could threaten to attack American shipping. This is a threat that Finland could execute.
A threat must also be credible, meaning that the threatened nation has to believe that the threatening nation will go through with it. In the above example, the US might feel that Finland would have far more to lose by attacking American shipping than it would be willing to pay. Since most threats leave both parties worse off, without other action most threats are considered empty.
There are two main ways to make a threat credible. One is to stake a state's or leader's reputation on it. This works best for democratic states, where the leaders have to go before the people later for reelection. The other main way is to "sink costs" into the threat. In the above example, if Finland were to put a hospital and a military base in Sweden, then it has strong reason to defend Sweden against US aggression. Another way is to forego the concept of rationality entirely and simply claim that you're so emotionally beholden to a certain stake that you just might retaliate excessively even in the face of a limited challenge to interests that wouldn't warrant such a response from a rational actor. Richard Nixon (unsuccessfully) tried to pull such a bluff off when threatening nuclear retaliation against North Vietnam, dubbing it the "Madman theory".
 Current deterrents
During the Cold War, deterrence theory mostly focused on the antagonistic relationship between the two superpowers - a concept called Mutually Assured Destruction (note the nice acronym), based on the enormous nuclear arsenals both maintained and the resulting impossibility of a successful first strike in the face of massive redundancy in delivery systems, was expected to ensure a rough balance of power and keep either of them from challenging the status quo.
However, nowadays, very few states are worried about direct nuclear attack. The US is mostly worried about extended deterrence, which is deterring attacks on allies, although there's not a lot of capable potential challengers left. The US presence in South Korea, originally designed to keep the Korean War from flaring up again, used to be a prime example of deterrence with conventional forces (nowadays, it's more about affirming a US commitment to East Asia and preventing arms races between other regional powers, which is also a form of deterrence, but not so much against a single specific threat). The continued presence of US forces in the Persian Gulf after the conclusion of the 1990-91 Gulf War was another example. The US also engages in "dual deterrence", in which it uses vaguely worded pledges of assurance to keep two sides for each doing things to lead to war. An example of this is the way the US pledges to aid Taiwan in its self-defense if attacked by China, but keep the form of such aid ambiguous and explicitly denouncing any efforts by Taiwan to declare full independence. This is (hopefully) keeping China from attacking and keeping Taiwan from declaring independence at the same time.
Russia's behavior towards some of its former fellow Soviet Republics and satellite states willing to turn to the West can be seen as an example of deterrence as well, although one that's more focused on economic rather than outright military threats (Russia arguably engaged in the latter vis-a-vis Georgia when trying to discourage them from challenging the quasi-independence of South Ossetia and Abkhazia, although this was apparently not taken seriously and resulted in a brief war in 2008). Mostly, nations such as Ukraine and Belarus have seen their gas prices soar as a result of taking actions that displease Moscow.
 Crisis of deterrence
While deterrence is theoretically about preventing war, it can inadvertently lead to war. Particularly because democratic nations feel a need to maintain a reputation, rash leaders can sometimes commit themselves to strong positions in the hopes that the other will back down, and end up going into war. This is usually due to a miscalculation of intents. The best example of this is the Korean War, where Mao, Stalin, Truman, and the elder Kim all misread each other and the levels of commitment they were willing to give to their respective sides.
Apart from this, a credible threat (especially a nuclear one) was once thought to require a high level of military readiness, very short response times and substantial autonomy for lower-level commanders in order to prevent a potentially successful enemy decapitation strike from rendering the nation unable to respond. Needless to say, all of these factors also increase the risk of an all-out war breaking out due to overreactions and misjudgement.
In the 1960s, however, the focus shifted instead to survivability of a nuclear deterrent, so that any given actor would be able to retaliate with a crippling "second strike". This removed the need to maintain military readiness or a short response time, as it could be put into play even days after the initial strike. This is why more effort has traditionally been put into protecting nuclear weapons, rather than population centers: if an enemy believes he cannot destroy a satisfactorily high percentage of your nuclear arsenal, he will presumably not attack you with nuclear weapons (but see next paragraph, as well as the first of this section), and so protecting your second-strike mechanisms should serve to ensure the safety of your population centers.
Another challenge to deterrence theory comes in the form of questioning the underlying rational actor model, especially since outright irrational behaviour can sometimes be advantageous: since an all-out nuclear war between the two superpowers would have left no survivors (at least not in the form of organized societies), there simply are no stakes that would warrant actually implementing the threat and facing one's own destruction at the hands of the challenger in turn. This problem is exacerbated in cases of extended deterrence, when a nation tries to cover allies overseas (as the US did in the Cold War). Convincing the opponent that you would implement a threat, even if it would not be rational to do so, can be a huge strategic advantage and cause him to "blink" first; this situation can be demonstrated by resorting to a chicken-type game theoretic model - think of the car race scene in "Rebel Without a Cause".
Some thinkers, notably Schelling, have tried to compensate with a schematic that draws on the notion of causing threats that leave things to chance. While no rational state would ever really threaten annihilation over an unimportant ally, states can easily threaten to increase the chance of war. This is exactly what Kennedy did by launching a quarantine of Cuba, for instance. Moreover, they have incorporated the ideas of "chicken" (in a fully game-theoretic format) into their theories.
 Successful deterrence
Kennedy's face off with Khrushchev over the Cuban Missile Crisis in 1962 was a demonstration of successful deterrence.
 Failures of deterrence
The most high profile failure of deterrence ever was the above-mentioned Korean War, when a weak US commitment to South Korea led to a North Korean invasion, and a seemingly incredible threat by China was ignored by the US leading to Chinese intervention. However, both World War I and World War II can be considered failures of deterrence as well. In World War I, the point of the Anglo-Russo-French Entente was to keep Germany from attacking any of the three, and Russian support of Serbia was intended to prevent anyone from attacking it. In World War II, the pledges of support for Poland were supposed to deter Germany from attacking, but failed. Some[who?] have questioned this, instead suggesting that the pledges to Poland were really a means of buying time to build strength for the war that was inevitable.
- Patrick M. Morgan. 1977, 1983. Deterrence: A Conceptual Analysis, 2nd Ed.. Beverly Hills, CA: Sage Publications.