The potential and the limitations of the Java language for graphics and visualisation research are evaluated. Evidence is offered that Java is adequate for high-performance computing, and still has the added benefit of the portability of its compiled byte code over a wide range of platforms. It is demonstrated that, contrary to popular belief, there is an efficient way for reading and writing individual pixels. It is also demonstrated that graphics primitives can be drawn in one memory access per pixel even without pointer arithmetics. Just-in-time and hot-spot compiler technologies considerably improve the speed of the executable code, though some penalty remains for dynamic binding. Some inconsistencies and flaws in the design and implementations of the Java class libraries are noted. In general, however, Java offers both portability and adequate performance of the executable code, and also increased productivity due to its extensive class libraries, therefore it should be a serious consideration for researchers, developers and educators in graphics and visualisation.