We have applied the standard wavelet and the adaptive wavelet transform algorithms to the record of the Central England Temperature (CET) from 1659--1990. Peaks in the CET spectra include 7.5¿1.0 yr, 14.4¿1.0 yr, 23.5¿2.0 yr, as well as a previously unreported variation at 102¿15 yr. Our wavelet analysis of CET agrees with previous results from Singular Spectrum Analysis (SSA) by Plaut et al. [1995> and gives additional results of variability on longer timescales. The interdecadal and century-scale variability in CET is strongly dependent on the interval of analysis. Estimates of a data trend are also shown to be sensitive to the cutoff timescale of the filter. A cooling of ≈0.3 ¿C during 1659--1720 is found relative to the temperatures during the 1800s. The complex time dependence of the actual data cautions against using model-derived representations of natural variability on such long timescales.¿ 1997 American Geophysical Union