So you have decided to learn more about performing studies either to increase your knowledge base or for caring out a study of your own. One of the many elements which comprise a study is validity. Validity sets standards to which a person can judge research.
This presentation deals with the subcategory of construct validity. Simply stated, construct validity asks, "did what I do (the treatment or program), do what I thought it would?". Did the treatment or program and its influence on the outcome actually reflect how I constructed (theorized) the relationship?
Now on to the topic at hand:
The MTMM is a validational process, developed by Campbell and Fiske in 1959, that gives specific standards to diagnose construct validity. This technique improves upon the previous philosophical nomological network foundation by including methodology. As the name implies there are three elements to the MTMM:
Multimethod - uses various methods in measuring those traits.
Matrix - a table comprised of correlations of various methods and concepts.
MTMM relies on two interrelated subcategories of construct validity, convergent and
discriminent validity.
For example suppose you had a construct of boot size and wanted to predict it by measuring foot size, sneaker size, height, and shoe size.
You would then calculate the correlations of each item/trait (sneaker size, height, . . ) with
the construct (boot size) 
Foot, sneaker, and shoe size are quite related to predicting boot size and would each lead to high correlations with boot size. In this case the correlation may be close to one indicating a "perfect" relationship between the two items. The item of height may not be as closely related therefore possibly giving a lower correlation in comparison to the other predictors. In this case though, all the measures result in high correlations, therefore, convergent validity is confirmed.
On the other hand suppose you had the two constructs: boot size and hair color. If you measured boot size with foot and shoe size and measured hair color by genes and eye color the model would look as follows:
As foot size and eye color are not related, the correlations between them will be low.
Correlations near zero indicate that items are not very related to each other, which in this case is expected. The fact that all the correlations are low indicates discriminent validity.
The numerical value of a "high" or "low" correlation is relative to the situation and requires the researcher to make judgments.
