CSP is a well established process algebra that provides comprehensive theoretical and practical support for refinement-based design and verification of systems. Recently, a testing theory for CSP has also been presented. In this paper, we explore the problem of testing from a CSP specification when observations are made by a set of distributed testers. We build on previous work on input-output transition systems, but the use of CSP leads to significant differences, since some of its conformance (refinement) relations consider failures as well as traces. In addition, we allow events to be observed by more than one tester. We show how the CSP notions of refinement can be adapted to distributed testing. We consider two contexts: when the testers are entirely independent and when they can cooperate. Finally, we give some preliminary results on test-case generation and the use of coordination messages.