Operational underwater noise emitted at 8 m s-1 by a 550 kW WindWorld wind-turbine was recorded from the sea and modified to simulate a 2 MW wind-turbine. The sound was replayed from an audio CD through a car CD-player and a J-13 transducer. The maximum sound energy was emitted between 30 and 800 Hz with peak source levels of 128 dB (re 1 µPa2 Hz-1 at 1 m) at 80 and 160 Hz (1/3-octave centre frequencies). This simulated 2 MW wind-turbine noise was played back on calm days (<0 Beaufort) to free-ranging harbour porpoises Phocoena phocoena and harbour seals Phoca vitulina in Fortune Channel, Vancouver Island, Canada. Swimming tracks of porpoises and surfacings of seals were recorded with an electronic theodolite situated on a clifftop 14 m above sea level. Echolocation activity of harbour porpoises close to the sound source was recorded simultaneously via an electronic click detector placed below the transducer. In total we tracked 375 porpoise groups and 157 seals during play-back experiments, and 380 porpoise groups and 141 surfacing seals during controls. Both species showed a distinct reaction to wind-turbine noise. Surfacings in harbour seals were recorded at larger distances from the sound source (median< 284 vs 239 m during controls; p< 0.008, Kolmogorov-Smirnov test) and closest approaches increased from a median of 120 to 182 m (p < 0.001) in harbour porpoises. Furthermore, the number of time intervals during which porpoise echolocation clicks were detected increased by a factor of 2 when the sound source was active (19.6% of all 1 min intervals as opposed to 8.4% of all intervals during controls; p < 0.001).These results show that harbour porpoises and harbour seals are able to detect the low-frequency sound generated by offshore wind-turbines. Controlled exposure experiments such as the one described here are a first step to assess the impact on marine mammals of the new offshore wind-turbine industry.