In saying that imagination is the primary aspect of true understanding, an example might be helpful. When I was working on my book Psychiatry and the Human Condition:
I decided not to include any theories about illnesses which I had been unable to confirm by imagination.
Imagination was not the term I used at the time; I instead called it phenomenology and introspection. Phenomenology is the term for the subjective description of psychological symptoms - e.g. what it is like to experience an hallucination, an endogenously depressed mood etc.
So, in my discussions of psychiatric phenomena and the effects of psychotropic drugs, I only wrote about theories which I had been able to enact imaginatively in myself - by examining my inner state and extrapolating from personal experience.
For example, in discussing mania I drew on experiences of staying up all night - when working as a doctor on call, or on a transatlantic flight - and the different stages about how it felt. For depression I had experiences of glandular fever and influenza, and recovering from them. For psychotic states I had memories of dreams, and the brief hypnagogic hallucinations when dropping off to sleep.
For psychotropic drugs, I had the 'advantage' of having taken many of the therapeutic classes of agents as attempted (and failed) curative or preventive treatments for migraine. Since I never found an effective preventive agent, I went through quite a large number of therapeutic trials before giving up. In addition I had for several years been very attentive to the effects of any drugs I was taking for anything: for example the antihistamines taken for treating Hay Fever, or introspectively studying the effects of simple pain killers, and how they got rid of pain.
So I had a kind of 'library' of knowledge about how drugs affect the way I feel, and this - combined with scientific knowledge about drugs and their effects - meant that I had a basis for understanding what drugs might to to psychiatric illness.
This amounted to an imaginative knowledge of psychiatric phenomena - and clarifies that imagination can be understood in terms of looking-within and becoming aware of subjective feelings.
The value of imagination was both at the beginning and at the end of theorizing. In the first place, phenomenology provided the primary basis of theories of how diseases may be caused, what is their essence, and of drug actions.
Then, after the theory was developed and elaborated and tested against the existing scientific literature using normal 'scientific' knowledge; phenomenology also provided a final test of plausibility and adequacy - could I imagine, could I experience within myself, the hypothetical cause or illness or treatment? If I could enact the proposed mechanism in imaginative experience, then the hypothesis would be allowed to stand - but if I could not enact them, then I would eliminate the hypothesis.
For example, the book was delayed for a few months by my initial inability to imagine experiencing mania - and this was helped by discussing mania empathically with someone who had experienced it and remembered the experience (by 'empathically', I mean in such a way that I could feel and experience what that person felt). Only when I could imagine myself as manic, and then imagine how that mania might be treated - how the manic inner state could be alleviated by inner change, could I finish the book.
This, then, is an actual example of how imagination can work - in a scientific context. Of course, when I wrote the book I mostly left-out the subjective and imaginative aspects - or used them only as illustrative examples. So the published Psychiatry and the Human Condition book is written and justified almost as if everything in it comes from the scientific literature and with reference to the experience of eminent psychiatrists and other objective and publicly available sources of information.
However, in reality, in terms of scientific creativity - the imagination was primary. It was primary in terms of generating hypotheses, and also exercised a veto in terms of testing hypotheses.