Context. Software testing is one of the crucial phases in software development life cycle (SDLC). Among the different manual testing methods in software testing, Exploratory testing (ET) uses no predefined test cases to detect defects. Objectives. The main objective of this study is to test the effectiveness of ET in detecting defects at different software test levels. The objective is achieved by formulating hypotheses, which are later tested for acceptance or rejection. Methods. Methods used in this thesis are literature review and experiment. Literature review is conducted to get in-depth knowledge on the topic of ET and to collect data relevant to ET. Experiment was performed to test hypotheses specific to the three different testing levels : unit , integration and system. Results. The experimental results showed that using ET did not find all the seeded defects at the three levels of unit, integration and system testing. The results were analyzed using statistical tests and interpreted with the help of bar graphs. Conclusions. We conclude that more research is required in generalizing the benefits of ET at different test levels. Particularly, a qualitative study to highlight factors responsible for the success and failure of ET is desirable. Also we encourage a replication of this experiment with subjects having a sound technical and domain knowledge.