Fundamental physics has traditionally been a part of the Euro-Atlantic culture. So the Japanese physicists may be viewed as relative newbies but they have already imprinted their skills onto our collective body of knowledge.
Hideki Yukawa became the first Japanese Nobel prize winner and, in some sense, he founded the Western theoretical physics in Japan as we know it. His description of the nuclear force - when the difference between the strong force and the weak force wasn't quite understood - was arguably the most important step in the physicists' understanding of short-range forces.
It helped the people to figure out that massive particles mediate short-range forces. Pions in his original model play this role when it comes to an approximate description of the strong interaction; W-bosons and Z-bosons generalize this logic and describe the weak interaction.
But I think it's right to mention three more older Japanese physicists who belong to the same league: Sin-Itiro Tomonaga (thanks, Robert!), Yoichiro Nambu, and Tamiaki Yoneya.
Sin-Itiro Tomonaga benefited from interactions with Werner Heisenberg and his group while in Leipzig, Germany in the 1920s - a decade before their countries would become close "axial" allies ;-) - but he became a powerful independent weapon who focused on the cancellation of divergences in quantum field theory. He discovered the old-fashioned renormalization independently of Schwinger. Tomonaga, Schwinger, and Feynman shared the 1965 physics Nobel prize.
Of course, Nambu joined two other Japanese physicists, Makoto Kobayashi (K) and Toshihide Maskawa (M), when they formed the trio of the 2008 Nobel prize winners.
Of course, many people had expected the Nobel-winning trio to reproduce the names behind the CKM matrix - we're used to it, after all. However, Nicola Cabibbo (C), while a very important physicist, wasn't quite in the same league as Nambu, I think. That's why I found the purely Japanese 2008 selection to be creative and fair. Unfortunately, Cabibbo died last Summer so he won't be able to get his own prize.
Nambu is a giant of theoretical physics. As a co-father of the Nambu-Goto action, he carved the cradle of string theory. As a key name behind the Nambu-Goldstone bosons, he was a critical man to uncover the secrets of the spontaneous symmetry breaking. And he is a co-father of the color, quarks' new charge, too.
Tamiaki Yoneya is two decades younger but I still included him among the great Japanese physicists of the older generation. Independently of Scherk and Schwarz, he realized that string theory automatically implies gravity as painted by general relativity. For many years, he would be working on various stringy versions of the uncertainty principle (for space and time etc.). As far as I can say, the details of this research never became conclusive - and holography has de facto showed that the ultimate correct logic of the inequalities is different - but his philosophy was important to shape the new "lore" of quantum gravity and string theory.
Let me get to two newer developments - linked to string theory: covariant string field theory and covariant matrix theory. It just happens that both of them are "covariant" versions of descriptions of string theory that happen to be working more reliably when they're not covariant. ;-)
Kyoto group and string field theory
When I was a high school student, I didn't have much access to serious books and articles about string theory. Although I knew everything that the Czech popular magazines had ever written about string theory, I couldn't really study it at the technical level. At most, I could read all (mostly misguided) Einstein's papers about the unified field theory - in the Pilsner Scientific Library where I have spent a lot of quality time. I didn't speak German well but you know, a dictionary helps and the actual key language were the equations. ;-)
That situation changed when I got to the Charles University in Prague at the end of 1992. It just happened that the first stringy book I borrowed from the library was a brown book called "Quantum String Theory" - if I remember well - with some proceedings from a 1987 conference. It reproduced several talks and two of them were accessible to me (and became very important in shaping my interests).
The first chapter of the book was about the light-cone superstring field theory. I think that it was written by Michael Green, John Schwarz, and Lars Brink (or Holger Nielsen?). I think that this early exposure has made it much more natural for me to learn the light-cone gauge versions of string theory when I began to study the subject with proper textbooks (especially Green-Schwarz-Witten).
The other chapter that greatly influenced me was one about the Kyoto group string field theory - designed by Hata, Itoh, Kugo, Kunitomo and Ogawa (HIKKO). It was meant to be analogous to Witten's cubic open string field theory but it was supposed to describe closed string interactions. The interaction vertex was able to merge two O-shaped strings into one 8-shaped one. In particular, the chapter also discussed the background-independent version of this string field theory whose action only has the cubic term, namely "A*A*A". I found that beautiful - for quite some time. The equation of motion is really "A*A = 0" which I found to be a perfect Ansatz for a theory of everything (on a T-shirt); I no longer think that this simplicity is the right cause for excitement.
Much of the maths in the HIKKO string field theory is analogous to Witten's cubic open string field theory - which I only encountered much later. Correspondingly, it took lots of time for me to understand that string theory doesn't necessarily have to be formulated as a string field theory. And it took an even longer time to understand why the Kyoto group string field theory is actually less consistent a description than the purely open string field theory by Witten.
Because of this incomplete information, I had believed for many years that the Japanese physicists had made the most advanced discoveries in string theory. It is no longer my exact belief today but the influence has been great.
The IKKT matrix model
A similar event got repeated four years later, at the end of 1996. Although I was monitoring the hep-th abstracts of the arXiv preprints on a daily basis, I had missed the BFSS matrix model paper in October 1996. What I couldn't miss, however, was a November 1996 paper by Vipul Periwal whose title was Matrices on a point as the theory of everything.
As far as I can say, the paper, while ambitious, wasn't right but the title revealed a good piece of marketing and it allowed me to re-discover the one-month-old BFSS matrix model which I began to investigate intensely. It turned out that in the decoupled world of Central Europe, I had mostly missed the duality revolution - focusing on (mostly free-fermionic) heterotic string phenomenology for years instead. However, the BFSS matrix model was a rather straightforward way to fill all those gaps. I don't want to discuss my own exciting adventures here.
But there was one more matrix model that was published soon, in December 1996, the IKKT matrix model by Nobuyuki Ishibashi, Hikaru Kawai, Yoshihisa Kitazawa, and Asato Tsuchiya. I would refer to it as the Japanese matrix model.
Well, it was rather a reproachful word than an expression of admiration but the model has been important because it was promising to become a covariant version of a BFSS-like map. However, the IKKT matrix model would probably be relevant as a non-perturbative (covariant) description of type IIB string theory instead of a non-perturbative (light-cone) description of M-theory offered by the matrix model.
Despite dozens or hundreds of hours I spent by checking this model and attempts to prove it - and reading of other people's papers about it - I am still undecided whether the model actually works or it is just hogwash that looks like something that is right but it is not right.
One thing is clear: Nathan Seiberg-like proof of this model remains non-existent and the papers written about the IKKT matrix model are much more sloppy and much less careful than papers written about the BFSS matrix model. Most of the people who think that the IKKT model is established at the same level as the BFSS matrix model simply have no clue what they're talking about.
I could tell you my strategies how to prove that the IKKT model is right if it is right - ways to reorganize all type IIB configurations into a small perturbation of a collection of N type IIB D-instantons by some "large" (and possibly complexified) SL(2,Z) U-duality transformations, replacing Seiberg's boost used to prove the BFSS matrix model. But it has never quite worked.
It still seems crazy to me that this very important question - whether the only semi-viable proposed covariant non-perturbative description of a superselection sector of string theory is right - hasn't been clearly settled by some good enough physicists and solid enough arguments. Of course, in the late 1990s, many people began to write papers on idiocies such as the anthropic principle so the number of people who could think hard about important but somewhat technically demanding problems - and I surely don't mean just this one - has decreased considerably.
Japan still hasn't managed to get to the top of theoretical physics but many very good people in the field are working in Japan, greetings to them. And I am confident that the average quality of the papers produced in the land of the rising Sun is much higher than the quality of the Chinese papers, for example - no offense to China, please, this is just a piece of reality.
Of course, Japan also has some serious experimental particle physics. The KEK accelerator center is known to many people because SPIRES offers the scanned version of all particle physics papers that were digitized by the KEK library; and Kamiokande has done lots of work in the search for proton decay and in neutrino physics.