The effect of beech bark disease on tree growth was tracked using paired resistant and susceptible American beech trees in two locations in Maine. Within each site, the paired trees were chosen in close proximity and with similar morphological characteristics (e.g. stem diameter and crown class) to minimize environment effects in subsequent analysis. A Kalman filter approach was employed to analyse the yearly time-dependent mean differences between paired susceptible and resistant tree-ring widths using simple structural time series models in state space form. On one site, under the influence of a moderate maritime climate, stand dynamics is hypothesized to account for the 34 year difference in onset of decline of trees in codominant, versus those in the intermediate crown classes. The harsher winter conditions associated with the second, more northerly site and known to limit the insect component of the disease complex, are hypothesized to be more of a factor in the close (six year) difference in decline onset between the two crown classes on this site. Some strengths and cautions in the Kalman filter approach are discussed in relation to the analysis of time-dependent trends in tree-ring series.