We present a new method for determining the rates from a network of regional tide gauge data. The method involves using a model for the sea level observations (annual averages of relative sea level) that explicitly includes a term for the interannual sealevel variations, which are assumed to be constant for all sites in the network. The resulting simultaneous analysis of all the sea level observations yields sea level rates which are biased by the unknown temporal slope of the interannual variations. By differencing sea level rate estimates with respect to one site, site-referenced sea level rates are calculated which are free from this bias. The differencing procedure also removes any signal due to a common sea level rate, such as might be associated with eustatic sea level rise. The resulting site-referenced sea level rates therefore in principle reflect only the relative vertical crustal motions of the tide gauge sites with respect to the geoid. These site-referenced sea level rates represent a potentially more accurate data set than have previously been used for sea level studies. We have used these rates to investigate glacial isostatic adjustment (GIA) in Fennoscandia. Specifically, we have used the site-referenced sea level rates within the Baltic Sea to estimate adjustments to the ice history and Earth model which are used to calculate vertical rates of adjustment due to GIA, a method never before used for Fennoscandia. The parameters selected were lithospheric thickness, the viscosities of (assumed isoviscous) upper and lower mantles, and a Fennoscandian ice thickness scaling parameter. We present the results of this estimation procedure for a specific test ice model. We also find evidence that the sea level record for the site nearest the center of uplift, Furu¿grund, may be contaminated by systematic errors of unknown origin. Data from this site have previously been used to determine a rate for the maximum uplift that we find is perhaps 1 mm yr-1, or ~10%, too large. ¿ 1999 American Geophysical Union