We use an inversion method based on stopping phases (Imanishi and Takeo, 2002) to estimate the source dimension, ellipticity, and rupture velocity of microearthquakes and investigate the scaling relationships between source parameters. We studied 25 earthquakes, ranging in size from M 1.3 to M 2.7, that occurred between May and August 1999 at the western Nagano prefecture, Japan, which is characterized by a high rate of shallow earthquakes. The data consist of seismograms recorded in an 800-m borehole and at 46 surface and 2 shallow borehole seismic stations whose spacing is a few kilometers. These data were recorded with a sampling frequency of 10 kHz. In particular, the 800-m-borehole data provide a wide frequency bandwidth with greatly reduced ground noise and coda wave amplitudes compared with surface recordings. High-frequency stopping phases appear in the body waves in Hilbert transform pairs and are readily detected on seismograms recorded in the 800-m borehole. After correcting both borehole and surface data for attenuation, we also measure the rise time, which is defined as the interval from the arrival time of the direct wave to the timing of the maximum amplitude in the displacement pulse. The differential time of the stopping phases and the rise times were used to obtain source parameters. We found that several microearthquakes propagated unilaterally, suggesting that all microearthquakes cannot be modeled as a simple circular crack model. Static stress drops range from approximately 0.1 to 2 MPa and do not vary with seismic moment. It seems that the breakdown in stress drop scaling seen in previous studies using surface data is simply an artifact of attenuation in the crust. The average value of rupture velocity does not depend on earthquake size and is similar to those reported for moderate and large earthquakes. It is likely that earthquakes are self-similar over a wide range of earthquake size and that the dynamics of small and large earthquakes are similar.