Boundary manikins, the concept of creating statistically extreme cases to accommodate a big part of the less extreme population has been known for decades. Despite this, many ergonomics simulations are done with few human models. This fact can be explained by the time consuming processes when working with many manikins in current digital human modelling (DHM) tools, but may also be a result of difficulty to understand how these manikins are calculated and defined. This paper focuses on the method of defining boundary manikins and how that functionality can be integrated into a DHM tool. Examples of boundary case methods in the literature often use Principal Component Analysis (PCA) which makes it possible to reduce the dimensions of the problem without much loss of the variance of the analysed data. Using PCA often demands some extent of manual analysis at the critical stage of reducing dimensions. This paper will explain a similar methodology for ceating boundary manikins from any number of variables, i.e. anthropometric variables chosen as key measurements. This method of creating a group of manikins is intended to be used in an automatic simulation feature in the IMMA software being developed in the associated research project. By using the method, a confidence region in the standardized space is created from eigenvectors and scaled eigenvalues of a correlation matrix. Boundary manikins are chosen at the ends of the axes of the enclosing confidence region, and one manikin of mean values is also added to the group of manikins. In the method presented here, the number of manikins created depends directly on the number of variables, which lead to the fact that the decision making of which key measurements to consider has to be done carefully to not create an overwhelming number of manikins. In comparison with one method using PCA, the method presented in this paper creates more manikins with a bigger difference in the max and min values of the chosen key measurements. If a limited number of cases are of crucial interest, then using PCA to reduce the dimensions of the problem is a good method to use. But if it is possible to create automated simulations the limitation of the number of manikins might not be so important. This will, though, depend heavily on the speed of the automated simulations.