Most of biological and medical investigations choose to involve either male or female (rarely both) subjects to study.
To merely include subjects from one gender (particularly males in rodent models in many studies) and to avoid populations of mixed genders in separate groups is almost a very well-established approach in experimental design of most of the studies in neurosciences, brain and cognitive sciences, cancer research, immunology and behavioral psychology, etc.
Nevertheless, I think that such an approach is often disputatious and may be inaccurate (though statistically solid only for those experimental groups), in that the real world population is not segregated, and research should truly reflect on the statistically unbiased real demographics.
Do you think that to avoid statistically unreliable variances due to biological gender differences in a mixed study (not mixed-gender groups), and thus to choose subjects from only one gender (male or female) would result in biased, unreliable or non-replicable studies?