We observed six components of ground rotational and translational motions in a near-field region during an earthquake swarm in April 1998 offshore Ito, Izu Peninsula, Japan. To check the reliability of our observation using a MotionPack sensor, we installed an 8301F sensor, which is an inertial angular displacement sensor with a higher sensitivity than MotionPack. The noise level of the MotionPack is much higher than that of the 8301F, but the waveforms of these signal components are quite similar to each other, suggesting that the MotionPack records rotational rate correctly but with higher noise. Our observation made known a linear correlation between the maximum rotational displacements around the vertical axis and the maximum translational velocities. The waveforms of rotational motion around the vertical axis are analogous to those of translational horizontal velocities, suggesting the linear correlation between maximum values. These rotational motions are much larger than those calculated by array data at the San Andreas fault. This large discrepancy might be explained by one or more of the following effects: the difference in the spatial scale of rotational motion by single gyro measurement and by array observation, the effect of the precipitous topography at the station of the offshore Ito swarm, and the difference of the degree of maturation between the San Andreas fault and the swarm regions of offshore Ito.