Approximately 1000 people were killed in the collapse of buildings in Istanbul, Turkey, during the 17 August 1999 I??zmit earthquake, whose epicenter was roughly 90 km east of the city. Most of the fatalities and damage occurred in the suburb of Avcilar that is 20 km further west of the epicenter than the city proper. To investigate this pattern of damage, the U.S. Geological Survey, in cooperation with the Kandilli Observatory and Earthquake Research Institute (KOERI), deployed portable digital seismographs at seven free-field sites in western Istanbul, to record aftershocks during the period from 24 August to 2 September. The primary objective of this deployment was to study the site effects by comparing the aftershock ground motions recorded at sites inside and outside the damaged area, and to correlate site effects with the distribution of the damaged buildings. In addition to using weak-motion data, mainshock and aftershock acceleration records from the KOERI permanent strong-motion array were also used in estimating the site effects. Site effects were estimated using S waves from both types of records. For the weak-motion data set, 22 events were selected according to the criteria of signal-to-noise ratio (S/N ratio) and the number of stations recording the same event. The magnitudes of these events ranged from 3.0 to 5.2. The acceleration data set consisted of 12 events with magnitudes ranging from 4.3 to 5.8 and included two mainshock events. Results show that the amplifying frequency band is, in general, less than 4 Hz, and the physical properties of the geologic materials are capable of amplifying the motions by a factor of 5-10. In this frequency band, there is a good agreement among the spectral ratios obtained from the two mainshocks and their aftershocks. The damage pattern for the 17 August I??zmit earthquake is determined by several factors. However, our study suggests that the site effects in Avcilar played an important role in contributing to the damage.