Anthropogenic activities are increasing in the Arctic posing a threat to species with high seasonal site-fidelity, such as the narwhal Monodon monoceros. In this controlled sound exposure study, six narwhals were live-captured and instrumented with animal-borne tags providing movement and behavioural data, and exposed to concurrent ship noise and airgun pulses. All narwhals reacted to sound exposure by reduced buzzing rates, where the response was dependent on the magnitude of exposure defined as 1/distance to ship. Halving of buzzing rate, compared with undisturbed behaviour, and cessation of foraging occurred at 12 and ~7-8 km from the ship, respectively. The effect of exposure could be detected > 40 km from the ship. At distances > 5 km, the received high-frequency cetacean weighted sound exposure levels were below background noise indicating sensitivity of narwhals towards sound disturbance and demonstrating their ability to detect signals embedded in noise. Further studies are needed to evaluate the energetic costs of disrupted foraging due to sustained disturbance but the observed sensitivity should be considered in the management of anthropogenic activities in the Arctic. The results of this study emphasize the importance of controlled sound exposure studies in the wild to explore the auditory capabilities of odontocetes. Methods Six male narwhals were live-captured in August 2018 in the Scoresby Sound fjord system in East Greenland. The data were collected using animal-borne AcousondeTM acoustic and orientation recorders and backpack FastLoc GPS-receivers (Wildlife Computers (Redmond, Seattle, WA, USA) collecting an unrestricted number of FastLoc snapshots through August 2018. Acousondes were set to collect triaxial acceleration and orientation (sf 100 Hz), depth (sf 10 Hz), and acoustics. Acoustics was sampled continuously with a 25 811 Hz sampling rate (HTI-96-MIN hydrophone, nominal sensitivity -201 dB re 1 V / μ Pa, preamp gain 14 dB, an anti-aliasing filter with 3-dB reduction at 9.2 kHz and 22-dB reduction at 11.1 kHz, 16-bit resolution). The dataset has been processed as follows: Time-depth records were down-sampled to 1 Hz and time-synchronized with GPS positions. Additional GPS positions were created for each second between successive positions through linear interpolation. Buzzes were used as a proxy for foraging attempts and were detected using a custom-written detector (Matlab, The MathWorks, Inc., Natick, MA, USA) and verified manually. When the sound source and animal were within line of sight (determined visually from maps showing the positions of the ships and whales aligned in time), distance between the whale and the sound source was determined for each second. The dataset is comprised of 5 columns: 1) Ind: unique name for individual whale, 2) Buzz: 0 or 1. 1 indicates a start of buzz in that second, 0 indicates no buzz. Buzzes were detected as described above. 3) Seismik: 0 or 1. 1 indicates that the seimic airgun used in the experiment was active, 0 indicates times when it was not active. 4): LOS: short for line of sight. 0 or 1. 1 indcates times when the individual and the sound source were within line of sight, 0 indicates times when the whale and sound source were not within line of sight. 5) Dist_ship: distance between whale and sound source in meters (m). Calculated only when LOS=1. NA indicates times when the whale and sound source were not within line of sight.