Similarity theory can be applied within the roughness sublayer above forests with a greater confidence than has generally been believed. Previously, flux-gradient relationships based upon similarity theory have been reported to underestimate scalar fluxes in the roughness sublayer above forests by factors of 1.3 to 3.4, when compared to independent flux estimates (for example, eddy covariance or energy balance measurements). In our research, the validity of similarity theory was investigated between July 31 and October 25, 1995, above a mixed deciduous forest at Camp Borden, Ontario. Eddy covariance and flux-gradient measurements of carbon dioxide exchange were compared at four layers within the roughness sublayer, and both before and after leaf senescence. The pre-senescence results may be applied by researchers using similarity theory to measure the fluxes of scalars with a mainly crown-level exchange surface; the post-senescence results are applicable to the measurement of scalars with a ground-level exchange surface. The eddy covariance measurements used a Li-Cor infrared gas analyzer, and the flux-gradient (similarity theory) measurements featured a tunable diode laser Trace Gas Analysis System (TGAS). The TGAS yields a very high resolution measurement, and the carbon dioxide concentration difference was resolved to 300 parts per trillion by volume (ppt) based upon a half-hour sampling period. The measured enhancement factor (the ratio of independent flux estimates, in this case eddy covariance, to similarity theory fluxes) was smaller and occurred closer to the canopy than in most previous investigations of similarity theory. Very good agreement between the eddy covariance and similarity theory fluxes was measured between 1.9 and 2.2 canopy heights, and the mean enhancement factors measured before and after leaf senescence were 1.10 +/- 0.06 and 1.24 +/- 0.07, respectively. Larger discrepancies were measured closer to the canopy (1.2 to 1.4 canopy heights), and mean enhancement factors of 1.60 +/- 0.10 and 1.82 +/- 0.11 were measured before and after leaf senescence, respectively. Additional results suggest that researchers working over comparable forests may use similarity theory with reasonable confidence at measurement levels higher than 1.6 canopy heights, regardless of the location of the scalar exchange surface. Results from this research should encourage the use of the flux-gradient technique above forests, for example in cases when the fast-response instrumentation required for eddy covariance measurements is not available.