Medi Swift

Join us in shaping tomorrow's
medical landscape

Multitrait Multimethod Matrix (MTMM)

The field of social research has come a long way from the days when researchers sought pure measures and simple designs. The norm today is multiple measures, elaborate multimethod designs, and analysis strategies that acknowledge the complexities of the social world. A key element in the multimethod approach is the multitrait multimethod matrix (hereafter referred to as the MTMM). The MTMM was introduced by Campbell and Fiske in 1959 as an effective method for evaluating construct validity. It organizes evidence of convergent and discriminant validity into one matrix that can be used to compare the correlations of different measurements in a study.

In a typical MTMM, the X РY axes represent measures mt-mm of the same trait or concept, while the diagonals represent measures that are not related to each other either because they measure different traits or because they use different methods. The correlations along the diagonals, called monotrait-heteromethod correlations, provide data about convergent validity. These should be relatively large, indicating that the measures do a good job of measuring the same thing. The correlations across the diagonal, however, should not be very high or they may indicate a problem with the measurement. These are called heterotrait-monomethod correlations and are usually expected to be very low.

A typical MTMM also includes rows of correlations between different measures that share both the same trait and the same method, called homogeneous correlations. These are supposed to be moderate and unbiased, meaning that the differences in results between the measures are due to the same factors and that any bias in the results is equalized across the methods. These are the rows of correlations in the MTMM matrix that are shown in the figure below.

As more and more research in the social sciences takes a multimethod approach, it becomes necessary to develop new statistical techniques for analyzing these data sets. A new statistic has emerged in recent years that is especially useful in the multitrait multimethod matrix. This is the structural model and it allows for a more sophisticated evaluation of multitrait-multimethod data than the simpler models such as correlational analyses. It can be used to estimate the level of convergent and discriminant validity as well as to detect method variance, or halo effects. The MTMM has proved to be a very useful tool in validating a number of measurement instruments in the field of social work, including measures of criminality, prostitution, and drug activity collected by resident surveys, systematic social observations, and police calls for service. In addition, the MTMM can be applied to other types of measurement data, such as those collected from archival records. In the future, the MTMM will likely be an essential part of the multimethod approach in many areas of social work research. This will require the development of new methods for assessing construct validity, as well as new analysis approaches that can take advantage of these new statistics. The current generation of multitrait-multimethod studies is just beginning to be analyzed using these new tools.

Scroll to Top