A method and apparatus for receiving a media stream provided in a plurality of different, alternative encoded representations. Each representation comprises a series of discrete fragments and each fragment comprises a contiguous temporal segment of the stream. The method comprises: obtaining (240, 310) at least a part of a first fragment from a first one of the representations; obtaining (210, 320) at least a part of a second fragment from a second one of the representations, wherein the first and second fragments comprise temporal segments that overlap at least partially; decoding (250, 330) a first temporal segment of the media stream from the first fragment; and decoding (260, 350) a second, later temporal segment of the media stream from the second fragment. In this way, the method switches from receiving the stream in the first representation to receiving it in the second representation.