While Keirsey is most known for his Please Understand Me I (1978) and II (1998) books, inbetween them, he published Portraits of Temperament (1987).

You very rarely hear about this book. I gained a curiosity about it two years ago, when he published his latest book, Brains and Careers (2008), which reportedly mentions "roles of interaction" (
Initiator, Coworker, Contender and Responder), which is his name for the four other groupings associated with Linda Berens; the "Interaction Styles".
Some confusion began on the new book's site: http://www.keirsey.com/brains.aspx with the statement:

The idea that there are four differing temperaments, as defined in Please Understand Me (1978) has gotten into most of the nations in both West and East. However, even though the idea of four differing roles that people play in face-to-face interaction with one another was introduced in Please Understand Me II (1998), the idea did not catch on as well as did the idea of four temperaments."

I had none of the books (only getting them this year), and upon getting both PUM's, found nothing about interactive roles in them. So I figured it must have been "Portraits of Temperament", which I also thought I saw it attributed to somewhere.

So now, I 've finally gotten that one, and this is the book where he introduced the eight "intelligence variants" (last three letter combinations), formed by dividing the four temperaments by the new role-informative/directive factor. Only here, he calls them "RoleMessaging sub-types". Two of the names differ from the final names that would be used in PUMII. SFP's are "Players" instead of "Entertainers", and STJ's are "Monitors" instead of "Administrators". Some of the type names are different from both PUMI and PUMII. In one case, the INTP, which was Architect in both PUM's, is called "Designer". Some others follow PUMI, and some go on to be retained in PUMII.

Still, I can find no trace of the four new groups formed by factoring E/I directly with informative/directive. The book simply divided the four temperaments into the eight subtypes, and then dividing the eight by E/I into the 16 types (as you can see done on the Wikipedia article on the KTS).

So what I think that Brains and Careers page was referring to was the eight subtypes and directive/informative scale, which simply paved the way for the possibility of refactoring them into the four roles, even though they were not actually mentioned. (Unless the third edition I have was changed from earlier ones). So they apparently were first grouped like that in the new book.

That would mean that Berens did beat him at these "interaction" groups after all! (Unless he had been teaching them before the new book, or published them in some other previous work. Her model seemed to be complete around 2006). Seems funny, as I hear he is critical of her; that he would later on adopt that part of her model.

Portaits of Temperament also does give us a more detailed definition of directing and informing. As I've always said, it's not just literal issuing of statements in the form of information or direction, it's also (according to this book) defining the roles we have with others. If I'm directing someone, then I'm tending to take on a more authoritative role, and placing them in a submissive role. If I'm informing, I'm basically waiting for the other person to propose the role we are to play.

This makes it more clear that Informativeness is responsiveness (he even calls it this) or agreeableness, while Directiveness is less responsive. This makes it easier to match this to the people/task scale, or high or low "Wanted Inclusion". People who want less interaction from others will tend to want to define the roles, while those who want more will tend to allow the other person to assign the roles.

This book also was reportedly where he provided more of a rationale for making the NF Choleric, based on Galen being a physician covering "the negative side of temperament", describing their "tendency to occasionally fly into fits of rage". But this I do not see in there either. It was probably a paraphrase by the person who told me that, of PUMII p.119.

He also gives a bit more of his rationale for choosing the groups that he did, in contrast to the "error" Myers inherited from Jung. I'm hearing a lot of different parts of the story about how Myers-Briggs started out with a plain four temperament system (reportedly, Meditative, Spontaneous, Executive and social) and eventually merged it with Jung's functions through trying different letters. Keirsey had said in PUMII that Myers confused extraversion with S, and introversion with N. This probably involved the transition from the Meditative (I) and Spontaneous (both S and N) types.

So now in this book, he says that Myers' "Sensible Introverts" should have been "Sensible Judgers" (to include extroverts), and "Sensible extroverts" should have been "Sensible Perceptors" (to include introverts). Basically, SJ and SP instead of Si and Se. But he's making it sound like she actually associated these with the actual I/E dichotomy. Maybe she did, looking at how she previously did try to figure I/E in the four type groups.
(With that said, he then goes on to attribute his groups, to her. She eventually settled on ST, SF, NT, NF).

The Team Technology site's critique of his theory http://www.teamtechnology.co.uk/keirsey-analysis.html suggests he misunderstood Jung on the whole E=S thing.

BTW, has anyone read Brains and Careers? I'm curious about what he says about the interaction roles.