Because of a lack of suitable instruments, rotational ground motions have not been observed until the last decade. Rotational measurements in the near field of earthquakes in Japan (Takeo, 1998) indicate that rotational ground motions are many times larger than expected from the classical elasticity theory. After failing to obtain useful rotational ground motions (using similar rotational sensors as Takeo did), we deployed a far more sensitive rotational velocity sensor (R-1) at the HGSD station in eastern Taiwan. From 7 December 2004 to 12 November 2006, several hundreds of earthquakes were recorded during our Phase 1 operation. This was mostly a learning exercise to solve field operation problems; Phase 1 operations ended when our two R-1 sensors ceased to operate. A K2 R1 instrument was deployed in the spring of 2007 to start our Phase 2 operation. From 8 May 2007 to 17 February 2008, we ob- served 52 local earthquakes with good rotational velocity signals (with signal-to-noise ratio >∼5), together with excellent translational acceleration signals (with signal-to- noise ratio >∼10). Unfortunately, field operation was interrupted due to flooding of the HGSD station site in mid-February 2008; we just resumed normal operation in June 2008. This article reports our observations of rotational and translational ground motions made at the HGSD station so far. We concentrate on describing our instru- mentation and the data obtained from 52 local earthquakes during our Phase 2 opera- tion and present some very preliminary results.