Digital multimedia systems are creating many new opportunities for rapid access to content archives. In order to explore these collections using search, the content must be annotated with significant features. An important and often overlooked aspect o f human interpretation o f multimedia data is the affective dimension. The hypothesis o f this thesis is that affective labels o f content can be extracted automatically from within multimedia data streams, and that these can then be used for content-based retrieval and browsing. A novel system is presented for extracting affective features from video content and mapping it onto a set o f keywords with predetermined emotional interpretations. These labels are then used to demonstrate affect-based retrieval on a range o f feature films. Because o f the subjective nature o f the words people use to describe emotions, an approach towards an open vocabulary query system utilizing the electronic lexical database WordNet is also presented. This gives flexibility for search queries to be extended to include keywords without predetermined emotional interpretations using a word-similarity measure. The thesis presents the framework and design for the affectbased indexing and retrieval system along with experiments, analysis, and conclusions.