Some therapies involve various forms of imaginary reenactment, where you heal a trauma by first recalling the memory of it and then imagining how things could have gone differently. Sometimes the imagined alternative can be quite fantastical in nature, such as your current adult self traveling back in time to when you were a child and saving your child self from the bullies tormenting you. (Here by trauma I mean to also talk about “small-t trauma”, e.g. various painful experiences that might not be what we’d ordinarily call trauma, but are still a little unpleasant to think about, or have left some other kind of a negative effect on your psyche.)

In my experience, imaginary reenactment works, at least assuming that I’ve managed to get an emotional hold of what exactly in the memory it is that made it feel so unpleasant. (Did I feel like I was alone? Or inadequate? Or that I did something wrong? Etc.) Also assuming that the memory of the old trauma isn't so painful as to be completely overwhelming and leave no room to imagine any alternatives.

Here’s my current guess of how and when this works:

The basic process by which any emotional learning gets changed is memory reconsolidation. There’s a generalization that your mind has drawn about the meaning of some past event that feels true to you. E.g. “nobody helped me when I was in that situation, so nobody cares about my suffering”. If you can bring that felt truth to mind while also experiencing a contradictory belief – e.g. the belief that you have a friend who does care about you – as true at the same time, your brain will notice that it believes in two contradictory things at the same time, and will revise its beliefs to fix that inconsistency.

Often, this takes the form of concluding that what it considered to be a general truth isn’t the case after all – e.g. changing the previous assessment to “nobody helped me in that situation, but there are still people who care about me and who I can reach out to for help”.

Now, you can also imagine things that feel true, if they’re the kinds of things you feel could happen. For instance, maybe you have a friend who buys a lot of products from the Acme Corporation, and you then imagine your friend excitedly telling you about the Acme Super-Duper Toothbrush that they bought. Even if they have never done this, the imagined scene can still feel real because it involves the kind of a thing that your friend could do.

I suspect what’s going on in therapeutic reenactment is that you are imagining something that feels like it could have happened and thus serves as counterevidence for the emotional belief in your trauma, but the “could be true” is on an emotional or symbolic level rather than on the level of physical possibility.

So for example, suppose that I had a childhood experience where I was being picked on by bullies and nobody helped me. From this experience, my brain might form the generalization “nobody helped me, so nobody cares about my suffering”.

Now if I manage to recall this experience in such a way that I can feel empathy towards my past self, then the act of feeling that empathy now proves that someone does care. Then if I imagine a scene in which I travel back in time to when I was a child and I beat up my bullies, it doesn’t matter if the literal content is physically impossible. Because what matters is the emotional feeling of “someone cares so someone could have helped”, which is evaluated as true.

(My adult self caring about my child self doesn’t mean that my adult self could actually have helped my child self, but one person caring is enough to disprove the generalization of “nobody cares”. So then if at least one person cares, then that implies that there were also others who would have cared, and they would have helped if they’d known and had the opportunity to. More broadly, the reason why we draw generalizations from past experiences is to predict the future, so what really matters is knowing that it’s possible to get help from people in general, and that people don’t think that your suffering is intrinsically meaningless.)

That said, if I try to do this and I haven’t really gotten a good intuition of why the memory is so painful, it usually doesn’t work – the generalization that I have formed from the experience needs to be at least somewhat explicit. Otherwise I can’t experience that generalization as real (as is required for the memory reconsolidation process to work), nor can I find the right emotional flavor that I need to imagine for the new scene to count as counter-evidence for the generalization.