Fuzzy logic techniques have been widely used in civil and earthquake engineering applications in the past four decades. However, no thorough research studies were conducted to use them for deriving attenuation relationships for peak ground accelerations (PGA). This paper is an attempt to fill this gap by employing a fuzzy approach with fuzzy sets for earthquake magnitude and distance from source with the objective of proposing new ground motion attenuation models. Recent earthquake records from USA and Taiwan with magnitudes 5 or greater were used; and consisted of horizontal peak ground acceleration recorded on three different site conditions: rock, soil and soft soil. The use of Fuzzy models to quantify ground motion records, which are typically characterized by a high level of uncertainty, leads to a rational analytical tool capable of predicting accurate results. Testing of the fuzzy model with an independent data set confirmed its accuracy in predicting PGA values. (C) 2014 Elsevier Ltd. All rights reserved.