In this paper, we propose a collaborative brain-computer interface for the automatic discrimination of images containing specific targets. When a user looks at a stream of images that are displayed using the rapid serial visual presentation protocol, images containing targets elicit a P300 event-related potential that can be detected. This allows the images to be automatically labelled as targets and non-targets. While this is relatively unreliable with single users, by combining the evidence from groups of users, we obtained relative median improvements in the accuracy (assessed by the area under the receiver operating characteristic curve) of 8.2% and 14.3% for groups of 2 and 3 users, respectively.