The Reviewed Event Bulletin for 16 August 1997, produced by the prototype International Data Center (pIDC), reported a small seismic disturbance in the vicinity of the northern Russian test site at Novaya Zemlya (at about 02:11 UTC). Initial reports suggested that this disturbance was caused by a small nuclear explosion, whereas others identify the source as an earthquake using vertical-component S/P ratios. However, other authors show evidence that questions the validity of the S/P method for this region and conclude that the disturbance cannot be positively identified as an earthquake using seismological evidence. The apparent attenuation from the 16 August 1997 disturbance to SPITS (Spitsbergen, distance Δ = 1280 km, azimuth = 317°) is reported to be weak. Similar low attenuation is suggested from high-frequency observations at KEV (Kevo, Finland, Δ = 1130 km, azimuth = 268°). Thus, high-frequency P and S waves propagate efficiently across the Barents Sea, leading to the possibility that the radiation pattern of the source may be inferred. We show, using a grid-search method, that the high-frequency SV/SH amplitude ratios observed at SPITS and KEV (measured from the complex envelopes of the free-surface corrected three-component seismograms), combined with simple Pn signals (with positive first motion) at HFS (Hagfors, Sweden) and NORES (Norway), are consistent with a double-couple source, with a range of orientations. We compute synthetic seismograms for one such double-couple source oriented at σ = 255°, δ = 115°, ψ = 120° (similar to that reported for the 1 August 1986 Kara Sea earthquake). The synthetic Pn seismograms at HFS and NORES and the relative SV/SH amplitudes of the synthetic Sn seismograms at KEV and SPITS match those observed reasonably well. Our model also predicts (1) weak Pn at NRI (Norilsk, Russia), suggesting that the emergent Pn onset and low signal-to-noise ratio observed is related to the focal mechanism, and (2) the negative P first motion and large SH/SV ratio observed at AMD (Amderma, Δ = 360 km, azimuth = 152°). Thus, we conclude that the 16 August 1997 seismic disturbance was an earthquake beneath the Kara Sea and that identification is possible using only data that would normally be available to the pIDC (assuming KEV can be used as an alternative for the nearby station ARCES, which was not working at the time of the disturbance).