Theoretical predictions of eastern North American (ENA) ground motion parameters based on a stochastic model (Boore and Atkinson, 1987; Atkinson and Boore, 1990) are evaluated in light of recent data, including data from the 1988 Saguenay, Quebec earthquake. The evaluation is based on visual comparisons of predicted and observed ground motion amplitudes, and on regression analyses of the data. Data are consistent with the theoretical model on average, although high-frequency ground motions from the Saguenay earthquake are underpredicted. It is hypothesized that differences between the observations and the stochastic model predictions may be explained by the presence of two corner frequencies in the source spectrum. Any single earthquake may exhibit ground motions significantly higher or lower than predicted due to local or earthquake-specific effects not accounted for in predictions of ‘average’ motions.