The following is a translation from the Romanian "Pe scurt despre analiza tranzactionala" by Mircea Popescu.
Transactional analysis is regarded in some academic circles (especially those of psychoanalytic orientation) as a rather unserious concern, if not completely unscientific. Certainly, the particular talent of Eric Berne to speak in terms accessible to the public at large has created sufficient interest and a following of "practitioners", more or less in the way of science or reason. Nevertheless, I think that the discipline is intellectually recuperable, in the following terms:
There exist in principle two paths by which we can understand and represent the processes which have a place within the mind of man: Either by listening to the descriptions produced by the respective man, or else following the facts of the respective man. The fact that a good part of classical psychology uses the first source does not absolutely invalidate the the second.
Regarding phylogeny, it's clear that in its evolution from the most primitive ganglions to the neocortex, the brain was guided by the need to respond to stimuli. This is the function which created this organ. The fact that the types of stimuli which appeared first received solutions and responses of better quality than the types of stimuli that appeared later isn't surprising. For example, we have the capacity to observe light, so much so that we can follow with the free eye millions of stars in a summer sky. Just so, we have the capacity to follow the movement. Specifically because it seems unremarkable, the simple fact that we can catch a ball falling from the sky shows how well-adapted the human brain is. For all technology's worth, there doesn't exist even today a robot that can play basketball, even if there have been robots capable of playing chess for twenty years.
The light was here for a while, but the movement appeared on Earth all at once with living beings. Both are fairly old, like stimuli. Speech, on the other hand, is from a more recent time; it appeared more or less concurrently with people. It wouldn't be surprising then if the brain wasn't capable of responding to the stimulus of speech as quickly, as well as it responded to the stimulus of the basketball.
If a joke is told today, it serves nothing to laugh on Sunday, at church. If you're mocked, you can't go to the post office with a vicious retort. In the majority of social interactions, a spoken response has to come in a few seconds if it's going to be heard. What is asked of our brain, in the end? Nothing more and nothing less than the receipt of information about the sounds heard, to recompose these sounds in the sounds of language, to get words from them, and from these words phrases, and from these phrases to extract understanding, from which to then process into further understanding, from which to make other propositions which then break into words to be vocalized. In a few seconds. It's a lot.
It wouldn't be, then, all that surprising to discover that, faced with such pressure and fairly exaggerate demands, the brain, like any human thing, also commits some simplifications. Up to a point, it's obvious that people don't listen to all that is said by the person being answered when they're forming responses. Still more obvious is that people do not understand, before giving a response, what was said in fact by the person to whom the response is addressed. Sometimes, as though by a miracle, it happens that they hear and understand, but I imagine you'll agree that this'd be quite rare.
This is, evidently, a bad thing, but wherever there's bad there's yet good, and the merit of this theory is that it finds the good in this case. Namely, if the response doesn't necessarily consider what was said, what can it reveal other than the internal structure of he who responds? If you don't speak to the object, you speak about you yourself, as though from nothing, in a vacuum, incapable to produce anything.
If in truth the responses that don't respond to, address, nor take into consideration stimuli describe the respondent, there would have to be then, theoretically speaking, specific modes in which different people fail to respond, recognizable through regular practice. This would be the first prediction a theory of transactional analysis would face on a scientific basis.
Practice confirms this first prediction, through so called "games". One of the most well-known is "Yes, but".
The one: I'd like to get rid of my belly.
The other: You've got to go to a gym.
The one: Yes, but I can't afford the membership fee.
The other: You have to go to work.
The one: Yes, but I'd have to wake up too early.
The other: You have to stop using the elevator.
The one: Yes, but my knees hurt.
The other: You have to change your diet.
The one: Yes, but I have a sensitive stomach and I can only eat certain things.
Obviously, you could continue this dialogue indefinitely, with minimal effort, and just so you could extend it to any given domain. If you get bored of this, you could also play "if it weren't for you", "why does this always happen to me", "look what you made me do", "look how hard I've tried", "I'm only trying to help", the list is long enough.
The next following prediction would be that people can, once they've read these things described, revise their own behavior in the sense of improving the quality of their communication with others; for the one part recognizing those cases in which the discussion transforms into simple transactions without notional content, simply satisfying some necessities of mechanical functioning, and for the other part eliminating from their own repertoire pseudo-responses that don't respond to anything in particular.
To the degree that this second prediction were satisfied in practice, transactional analysis would have to become a school of clinical psychology. To what degree this satisfaction actually happens, only you can say.