Software testing is a very important phase of software development. Testing is carried out to detect as many numbers of faults as possible in the software. Several models are developed to address this question. But none of them are valid at all times. Data on number of errors and their categories (according to severity) are available for each module during testing of particular software. Assuming that the modules are independent and the reliabilities of modules are unknown, a model has been developed and an attempt has been made to estimate the total number of each category of defects in the software which roughly gives a measure of the reliability of the software.