Framed building structures rely on the integrity of beam to column connections to avoid collapse in the event of damage to some part of the building. This requirement is often referred to as robustness and in UK practice may be assumed to be achieved (for certain categories of building) if the beam to column connections are designed to resist a specified tensile force. However, prior to the current work, no published data existed on the behaviour of connections subjected to rapid, non-cyclical loading to failure, and models of frame response to localised damage typically use either very simplified assumptions for connection behaviour (e.g. fully rigid against translation and rotation) or data from static connection tests. This paper describes experimental equipment which has been designed, fabricated and commissioned in order to facilitate this type of testing. The experimental test rig is capable of applying an axial tension load or a combination of moment and tension to a connection. The loading can be applied at quasi static or high-rate dynamic rates, with connections taken to failure in a few hundredths of a second. A methodology for analysis of the data, in particular in allowance for inertial effects in interpreting the load/moment data is presented, along with validation tests and examples of results from connection tests.