Put enough pressure on someone and they will often change their story, even in crucially important situations like police interrogations. The question is, how much pressure is "enough"? The Gudjonsson suggestibility scale helps people find out.
Gísli Hannes Gudjonsson was a professor of forensic psychology, and so was fascinated with the idea of memory, and how often specific memories can be changed. He found that the memory, or at least the confessed memory, is especially suggestible in individuals who don't have a high degree of intellectual ability, leading to a lot of people with mental disabilities confessing to crimes they did not commit. Gudjonsson wrote about this in "The relationship between confabulation and intellectual ability, memory, interrogative suggestibility and acquiescence."
But even if police interrogators and lawyers knew that some people could be coaxed into false confessions, how were they to separate the people who broke down and confessed from the people who broke down and agreed? To that end, Gudjonsson came up with the Gudjonsson suggestibility scale. It's a series of tests, some meant to test a person's memory under neutral conditions, and some – requiring acting ability from the interviewer – meant to see how easily the subject will abandon those memories.
The test measures two main factors, "yield" and "shift." Yield is the degree to which a person will simply agree with leading questions. If a person asks "Don't you think that X is a little too bossy to have that job," it's easier just to say "I guess," than "No, I don't agree." We've all done it, if only to avoid an argument with a stranger on a bus. Sometimes people do it when the stakes are higher.
Shift involves a person slightly modifying their answers when they see that the interrogator isn't pleased. Sometimes all it takes is repeating the question multiple times. Children are especially susceptible to this, as they've learned to take cues from the adults around them and assume that, if they are asked to do something again and again, they have done it wrong. But almost anyone who senses consistent displeasure will try to soften or qualify their answers, voicing them as a question, or bringing up the limitations of their assessment.
The Gudjonsson scale is not uncontroversial, but it has stood up to a few tests. Special shortened versions of the test turn up results that are consistent with the results of the long version of the test. Different testers rating the same subject will turn up approximately the same results. There are even consistent ways of distinguishing people who are faking suggestibility from those who are actually highly suggestible. At least the test is consistent with itself, and not a total Rorschach.
How well does it rate people's susceptibility in practical situations? That depends. Even assuming the test is totally accurate, different mental states, including lack of sleep, stress, fear, and intoxication, mess with a person's score. So a person who will stand up to testing in one situation, will break down in another.