However, giving an example of experiment that creates an impression (true or false) that it was badly done and badly interpreted does not really support your case that much… After all, such experiments are also done in all kinds of pseudoscience…
And it just so happens that a short search (that has failed to find anything else about the experiment you mentioned - not that I have tried very hard) has found some studies (not necessarily impeccable themselves) that seem to show that the conclusion you mentioned might have been wrong. Just look:
sciencedaily.com/releases/2009/12/091208132241.htm,
nature.com/nature/journal/v463/n7279/full/nature08711.html,
nature.com/nature/journal/v485/n7399/full/nature11136.html,
med.mcgill.ca/mjm/v06n01/v06p032/v06p032.pdf (this review article even has a chapter “Problems with previous studies”)…
And the problem that has already been mentioned (experimenting with women and making conclusions about men) is not the only one. Let’s look at the description you provided again:
Some other potential problems:
*]You need a representative sample. It is not enough to “inject testosterone in women” - it is necessary to select the women in the way that let’s us reach conclusions about some larger population.
*]You need a control group.
*]You need a good way to measure “aggression”. Some of the studies I cited used a “ultimatum game”, but it is by no means self-evident that such method actually measures “aggression” (understood “normally”).
Thus, well, are you sure you do not want a lower (but still rather honourable) title “protoscience” for Psychology…? 