The goals of this work are to review the Northern-Italy ground-motion prediction equations (GMPEs) for amplitude parameters and to propose new GMPEs for frequency content and duration parameters. Approximately 10,000 weak and strong waveforms have been collected merging information from different neighboring regional seismic networks operating in the last 30 yr throughout Northern Italy. New ground-motion models, calibrated for epicentral distances ≤100 km and for both local (ML) and moment magnitude (Mw), have been developed starting from a high quality dataset (624 waveforms) that consists of 82 selected earthquakes with ML and Mw up to 6.3 and 6.5, respectively. The vertical component and the maximum of the two horizontal components of motion have been considered, for both acceleration (peak ground horizontal acceleration [PGHA] and peak ground vertical acceleration [PGVA]) and velocity (peak ground horizontal velocity [PGHV] and peak ground vertical velocity [PGVV]) data. In order to make comparisons with the most commonly used prediction equations for the Italian territory (Sabetta and Pugliese, 1996 [hereafter, SP96] and Ambraseys et al. 2005a,b [hereafter, AM05]) the coefficients for acceleration response spectra (spectral horizontal acceleration [SHA] and spectral vertical acceleration [SVA]) and for pseudovelocity response spectra (pseudospectral horizontal velocity [PSHV] and pseudospectral vertical velocity [PSVV]) have been calculated for 12 periods ranging between 0.04 and 2 sec and for 14 periods ranging between 0.04 and 4 sec, respectively. Finally, empirical relations for Arias intensities (IA), Housner intensities (IH), and strong motion duration (DV) have also been calibrated. The site classification based on Eurocode (hereafter, EC8) classes has been used (ENV, 1998, 2002). The coefficients of the models have been determined using functional forms with an independent magnitude decay rate and applying the random effects model (Abrahamson and Youngs, 1992; Joyner and Boore, 1993) that allow the determination of the interevent, interstation, and record-to-record components of variance. The goodness of fit between observed and predicted values has been evaluated using the maximum likelihood approach as in Spudich et al. (1999). Comparing the proposed GMPEs with SP96 and AM05, it is possible to observe a faster decay of predicted ground motion, in particular for distances greater than 25 km and magnitudes higher than 5.0. The result is an improvement in fit of about one order of size for magnitudes spanning from 3.5 to 4.5.