All targets have a variety of descriptive characteristics (that is, SAM target attributes). When comparing one target with another, both similarities and differences will be found between the two. The correspondence numbers are one measure of the degree of similarity between any two sets of SAM data, and these numbers can be used to compare one target with another target, or a remote-viewing session with a target. The correspondence numbers are calculated as per Test One. Proportion A is the total matches between the session and the target as a proportion of the total number of target attributes. Proportion B is the total matches between the session and the target as a proportion of the total number of session entries (not target attributes as with proportion A). The average of proportions A and B is called the "correspondence number" for the session, and it is a general measure of the correspondence between the observed remote-viewing data and the actual target attributes. Again, correspondence numbers can also be calculated between any two targets to measure their degree of similarity.

Test three evaluates the correspondence numbers for each session. The better a remote-viewing session describes all of a target's characteristics, the higher will be the correspondence number between the session and the target. Used in this way, the correspondence number is called the "session/target" correspondence number. When correspondence numbers are calculated that compare one target with another, such numbers are called "target/target" correspondence numbers.

We want to do two things with these correspondence numbers. First we want to note the relative ranking of the session/target correspondence number for the remote-viewing session and its real target as compared with the session/target numbers for the session and other (bogus) targets in a pool of targets. If a session describes the actual target relatively well, then its correspondence number should be high relative to alternative correspondence numbers for bogus targets selected from a pool. Second, we want to compare the variation of both the session/target numbers and the target/target numbers with regard to the pool of targets. Since a pool of targets normally contains targets with a great variety of descriptive characteristics, comparing any given real target with other bogus targets will result in finding various collections of similarities across the comparisons. For example, the real target may have a mountain and a structure. Comparing this target with another target that has only a mountain will find the similarity in terms of the mountain but not in terms of the structure. Comparing the same real target with another target that has only a structure will find the similarity with respect to the structure, but not with respect to the mountain. Using a number of comparisons in this way across a pool of targets allows us to account for all or most of the real target's important characteristics. This returns us to wanting to compare the variation between the two sets of session/target and target/target correspondence numbers across the pool of targets as a means of evaluating the overall success of a remote-viewing session in capturing its real target's total set of attributes. When compared with other targets which in the aggregate contain many different attribute sets, both the remote-viewing session and its real target should have correspondence numbers that vary similarly. The correlation coefficient summarizes this relationship. The correlation coefficient can vary between -1 and 1. The closer its value is to 1, the more closely the remote-viewing session describes all of its real target's various characteristics.

To begin this comparison in test three, correspondence numbers between the remote-viewing data and all 13 targets that were chosen for the public demonstration of remote viewing are calculated and presented in a table. This allows for a direct comparison of correspondence numbers between the remote-viewing session and the real target as compared with those numbers involving the other targets in this small pool. An accurate session should have a correspondence number for the real target that has a relatively high ranking as compared with the correspondence numbers involving the other targets. The correlation coefficient for the session/target and target/target correspondence numbers is also calculated. A high correlation between the two sets of numbers indicates that the session data and the target attributes for the real target for the experiment are similar when compared with target attributes for other targets in the public demonstration pool.

In Part II of this test, correspondence numbers for the given remote-viewing session and all targets in a diverse pool of 240 SAM targets are calculated. Additionally, correspondence numbers calculated using the real target for the remote viewing experiment and all targets in the SAM pool are also calculated. If the remote-viewing session describes the real target well, then the two sets of correspondence numbers (that is, one comparing the session with the SAM pool, and the other comparing the real target with the SAM pool) should vary similarly. Since it is impractical to examine and compare each pair of correspondence numbers using this larger pool of targets as is done in Part I for this test, only the correlation coefficient for the two sets of correspondence numbers is calculated and presented.