This paper compares the effectiveness of three query expansion techniques: global analysis, local feedback, and local context analysis. The study shows that local analysis, particularly local context analysis, is more effective than global analysis in improving retrieval performance. Local context analysis, which combines global analysis techniques with local feedback, outperforms simple local feedback in terms of retrieval effectiveness and predictability. The paper evaluates these techniques on three collections: TREC3, TREC4, and WEST. Results show that local context analysis performs significantly better than Phrasefinder (a global analysis technique) on both TREC3 and TREC4. On TREC4, local context analysis using 100 passages is 23.5% better than the baseline, while Phrasefinder is only 3.4% better. On WEST, local context analysis also performs well, though not as effectively as on TREC collections. Local context analysis is more robust and efficient than global analysis. It is computationally practical and can be applied to interactive applications. It also handles proximity constraints better than Phrasefinder and uses frequent but potentially good expansion concepts. However, it may require more time to expand a query than Phrasefinder. Local feedback, while effective on TREC3, is sensitive to the number of documents used for feedback and performs poorly on WEST. It tends to hurt queries with poor performance and those with few relevant documents. In contrast, local context analysis is less sensitive to the number of passages used and performs better on queries with few relevant documents. The study concludes that local context analysis is a better query expansion technique than both global analysis and local feedback. Future work includes automatically determining the number of passages and concepts to use, and improving the concept selection metric for Phrasefinder.This paper compares the effectiveness of three query expansion techniques: global analysis, local feedback, and local context analysis. The study shows that local analysis, particularly local context analysis, is more effective than global analysis in improving retrieval performance. Local context analysis, which combines global analysis techniques with local feedback, outperforms simple local feedback in terms of retrieval effectiveness and predictability. The paper evaluates these techniques on three collections: TREC3, TREC4, and WEST. Results show that local context analysis performs significantly better than Phrasefinder (a global analysis technique) on both TREC3 and TREC4. On TREC4, local context analysis using 100 passages is 23.5% better than the baseline, while Phrasefinder is only 3.4% better. On WEST, local context analysis also performs well, though not as effectively as on TREC collections. Local context analysis is more robust and efficient than global analysis. It is computationally practical and can be applied to interactive applications. It also handles proximity constraints better than Phrasefinder and uses frequent but potentially good expansion concepts. However, it may require more time to expand a query than Phrasefinder. Local feedback, while effective on TREC3, is sensitive to the number of documents used for feedback and performs poorly on WEST. It tends to hurt queries with poor performance and those with few relevant documents. In contrast, local context analysis is less sensitive to the number of passages used and performs better on queries with few relevant documents. The study concludes that local context analysis is a better query expansion technique than both global analysis and local feedback. Future work includes automatically determining the number of passages and concepts to use, and improving the concept selection metric for Phrasefinder.