Affect Misattribution Procedure is a new implicit research method of social cognition which is built on the foundation of projection. In the typical AMP, a trial is made of four parts, a evaluative prime image, a blank screen, a target Chinese pictograph image and a pattern mask consisting of black and white “noise” which appears until the participant responded. The prime, the blank and the target are presented about 75~125ms. The duration of prime pictures, target words or picture and the stimulus onset asynchrony only influences the effect of AMP slightly. There are three fundamental assumptions of AMP. The first one is when people see beautiful or nice things, positive emotion would emerge in their mind, on the contrary negative emotion would come into being. Secondly, these emotion will be expressed in their behavior, that is to say it can be measured. Thirdly, people can not realize that their emotion influences their judgement. A researcher investigated the internal mechanism of AMP, and his experiments substantiated that AMP process was made of 3 parts: the activation of emotion, the attribution, and the guess. After the prime picture, the emotion of subjects is activated firstly. But subjects would then attribute the emotion to the target picture, if their emotion has been activated. If the emotion is not activated or the motion is inhibited, they would guess. The superiority of affect misattribution procedure includes high reliability, nice validity, ease of use, and large effect size. The highlighted feature of AMP is the resistance to warnings, even in the very explicit salient warning condition subjects can not correct their attempts. They still misattribute their emotion to the target pictures instead of prime pictures. AMP has the property of relativity, in other words it needs a contrast object. But researchers can use AMP to study single attitude object, using a neutral object as a basic line, such as numbers, grey quadrates and so on. Comparing with implicit association test, AMP has some shortages in the researches of specific impression to some attitude targets. For example, it can estimate the valence of the attitude objects, but it can not ascertain the attitude content, such as the property of the attitude objects. But now researchers developed new amendments of AMP, and these methods supplied these gaps. Such as adding context factor in the typical AMP, requiring participants judge the prime stimulus directly, using four points measuring scale instead of “good” or “bad” judgement were new evolutions of affect misattribution procedure. And in the recent researches, the new variant of AMP is semantic misattribution procedure, which is well worth researchers paying close attention to. And these variants will make more researchers use AMP to explore the implicit attitude to genders, races, political candidates, alcohol, cigarettes, romantic relations and so on. The other advantages and disadvantages are discussed.