Despite a lot of initial enthusiasm and more than three decades of research, cognitive endophenotypes for psychiatric disorders are still to be found.

Based on a literature review and on our own research, we will analyse the reasons and consequences of this failure to find useful cognitive endophenotypes.

Several commonly held ideas that proved to be over-optimistic, over-simplistic and finally false, have limited our ability to identify cognitive endophenotypes. Among those ideas, with deleterious methodological consequences, were the beliefs that neuro-cognitive validity is sufficient to ensure genetic validity, that cognitive measures and cognitive processes are equivalent and that cognitive processes have a simpler genetic architecture than psychiatric vulnerability. The perception of these initial errors modified our definition and expectations of cognitive endophenotypes and suggested ways to improve our chances to find them.

Several aspects of the study of cognitive endophenotypes demonstrated an initial excessive optimism, followed by disillusion and, now, a time for active search for realistic solutions. We will illustrate this process by an important feature for cognitive endophenotypes: the test-retest reliability. Although cognitive measures were initially considered stable, a systematic literature review revealed that most of them had problematic test-retest reliability. The use of such measures could lead to erroneous conclusions and limit their usefulness as cognitive endophenotypes.

Taking this parameter into consideration is important in selecting cognitive tests used to detect putative endophenotypes and in suggesting new approaches in the search for cognitive endophenotypes (for example the use of cognition questionnaires).