Capabilities of major Japanese user agents, three screen readers and one voice browser, were investigated with the following test files: W3C UAAG 1.0 Test Suite for HTML 4.01, an accessible PDF file, an accessible Flash file, and test files which test Japanese specific issues. Using the UAAG 1.0 Test Suite, 20 out of 48 Priority 1 checkpoints were met by all user agents, while all of the user agents failed to meet 11 of the checkpoints. Test results of all test files were assigned into three categories: capabilities satisfied by almost all user agents, capabilities not satisfied by any of the user agents, and capabilities that were satisfied by some of the user agents only. The test results indicated that 1) two major Japanese user agents do not have enough functions to navigate through a Web page using the structure information of the content, and 2) none of the user agents have enough functions to control multimedia and time-dependent interactions. These results provide an objective evidence to define the Japanese baseline, a set of technologies that a user agent is assumed to support, which is required in the WCAG 2.0 working draft. Accessibility responsibility between Web content and user agents is also determined by the current survey. Copyright 2006 ACM.