Abstract Whereas traditional databases manage onlydeterministic information, many applications that use databases involve uncertain data.This paper presents a Probabilistic Tree Data Base (ProTDB) to manage probabilistic data,represented in XML. Our approach differs from previous effortsto develop probabilistic relational systems in that we build a probabilistic XML database.This design is driven by application needs that involve data not readily amenable to a rela-tional representation. XML data poses several modeling challenges: due to its structure, dueto the possibility of uncertainty association at multiple granularities, and due to the possi-bility of missing and repeated sub-elements. We present a probabilistic XML model thataddresses all of these challenges. We devise an implementation of XML query operationsusing our probability model, and demonstrate the efficiency of our implementation experi-mentally. We have used ProTDB to manage data fromtwo application areas: protein chemistry data from the bioinformatics domain, and informa-tion extraction data obtained from the web using a natural language analysis system. Wepresent a brief case study of the latter to demonstrate the value of probabilistic XMLdata management.