Diversity and evenness indices were widely used in community ecology and biodiversity researches. However, shortage of statistic tests on these indices restricted their reliability. To develop statistic test methods on diversity is one of the focuses in biodiversity researches. In present study, some randomization tests on statistic significance of diversity and evenness indices, confidence interval of diversity and evenness, and randomization test on statistic significance of between-community differences were presented. Shannon-Wiener diversity index, Simpson diversity index, McIntosh diversity index, Berger-Parker diversity index, Hurlbert diversity index, Brillouin diversity index, and corresponding evenness indices are included in the randomization test procedure. The web-based computational software for the statistic tests, BiodiversityTest , which is comprised of seven Java classes and an HTML file, is developed. It can be run on various operational systems and java-enabled web browsers and, may read ODBC linked databases such as MS Access, Excel, FoxPro, dBASE, etc. Rice arthropod diversity (15 sampling sites, 125 arthropod species, 17 functional groups) was recorded on September,1996 in IRRI rice farm using RiceVac apparatus and bucket enclosure. The data were analysed using BiodiversityTest with Shannon-Wiener index and Berger-Parker index respectively, and the results showed that the changes of diversity and evenness can be effectively detected by these tests. The randomization tests will correct the possible wrong conclusions aroused in direct comparison of arthropod diversity which was used in most of the researches up to now. The development of randomization tests on biodiversity will provide a quantitative tool for stricter statistic comparison of biodiversity between communities and present an absolute criterion fordiversity measuring. BiodiversityTest will make the computation realistic and accessible on Internet.