Object relational database management systems (ORDBMS) are becoming more popular in storing and retrieving XML than native XML DBMS. In most ORDBMS, XML is stored as CLOB inside the relation. Efficient XML parsers a nd indexing techniques are used to retrieve the desired XML nodes. However, less attention is given to XML updates queries. With the upcoming standa rdization of XML updates queries, the current implementation of the lock granularity imposes a great limitation on the concurrency of parallel transactions. This motivat ed several experimental ORDBMS to shred the XML nodes across internal relations. This approach has also several drawbacks. In this paper, we propose an adaptive technique for selective shredding. It is based on existing database engines and takes the changes in the workload pattern into consideration. We an alyze the performance of our approach and compare it to the CLOB and the complete shredding approaches.