This paper describes an experimental investigation into the seismic response of concentrically braced steel frames (CBFs). Twelve shake table tests were performed on full-scale single storey frames, each containing a pair of identical brace members. The experimental programme examined the behaviour of brace members with four different square and rectangular hollow cross-sections and a range of gusset plate connection details. The aim of the experimental study was to determine the influence of brace and gusset plate properties on CBF response from serviceability to ultimate limit states, including collapse. Consequently, all test frames were subjected to three levels of seismic excitation: (i) low-level excitation to examine elastic frame response, (ii) medium-level excitation to examine brace buckling and yielding effects, and (iii) high-level excitation to induce brace fracture. A detailed set of data on the seismic response of CBFs with realistic brace members and connections were obtained from the tests. The experiments were conducted under representative dynamic response conditions as opposed to the conventional idealised quasi-static loading procedures employed in previous experimental investigations of CBF behaviour. The results faithfully capture the behaviour of brace-gusset plate test specimens with different non-dimensional brace slenderness, brace cross-section slenderness, connection types and gusset plate detailing. The response variables measured in each test included the shaking table and frame accelerations and displacements, brace elongation and axial force, and brace member and gusset plate strains. The experimental observations include elastic frame vibration properties, acceleration and drift demands, ultimate failure modes and ductility capacity. The brace-gusset plate test specimens remained elastic at low-level excitations, brace buckling and yielding occurred in all medium-level excitation tests, while specimens exhibited brace fracture under high-level excitation. Fracture did not occur in the gusset plate connections irrespective of whether these were designed using a conventional design method with a Standard Linear Clearance (SLC), or a balanced design with an Elliptical Clearance (EC). However, the balanced design approach showed more uniform distribution of plastic strains and led to higher brace ductility capacities when compared to the conventional design method. Based on the test results, available methods for predicting the ductility of bracing members are compared and assessed, and a number of considerations for design are highlighted and discussed.