https://www.utdallas.edu/~herve/abdi-WiresCS-mfa-2013.pdf

This one will help you in this matter.

Thank you for your answer! I came across some useful links as well, this one highlights the arguments against using factor analysis before clustering: http://ro.uow.edu.au/commpapers/774/

.

When you write that "(you) know (you) have four factors along which (you are) seeking to cluster the data", do you mean that you've run a PCA and found that the four first factors capture enough variance so that you can safely proceed with the analysis inside the four factors space, or do you mean that expert knowledge has evidenced four factors driving the data (in other studies, for instance) and that you seek to confirm such findings ?

In the first case, it would make sense to use the loadings, in the second case, it would not.

(This may be why your litterature review shows both approaches !)

.

Also, a very useful article on the pitfalls of consensus clustering :

http://www.nature.com/articles/srep06207

.

Thank you for your answer! I haven`t run a PCA, but the scales we used are established ones for the four factors. In a different study, I did a factor analysis for these and addiotional factors in the data and the scales converged pretty well. That is why I am not so worried that the data might not be driven by the four factors.

Thank you for the article - that seems a very informative read, I`ll consider the points mentioned there!