Simultaneous measurements have been made of all three components of the fluctuating temperature dissipation in the inner region of a fully developed turbulent boundary layer at a moderate Reynolds number. Measurements are made with a four‐wire arrangement which consists of two parallel vertical wires mounted a small distance upstream of two parallel horizontal wires. Each of the four wires is operated at very low current by a contant current anemometer and is sensitive to only the temperature fluctuation &THgr;. The separation between wires in each parallel pair is kept small, so that the differences between the outputs of each pair are reasonable approximations to ∂&THgr;/∂zand ∂&THgr;/&Lgr;y, the temperature derivatives in the transverse and vertical directions, respectively. The streamwise derivative ∂&THgr;/&Lgr;xwas obtained from the time derivative, through use of Taylor’s hypothesis. Mean square and spectral density measurements show that in the inner region local isotropy is not closely approximated [(∂&THgr;/∂z)2≳ (∂&THgr;/∂y)2≳ (∂&THgr;/∂x)2] and (∂&THgr;/∂x)2is richer in high frequency content than the other two components or the sum. The probability density of the sum &khgr;[= (∂&THgr;/∂x)2+(∂&THgr;/∂y)2+(∂&THgr;/∂z)2] has a lower skewness and flatness factor and is more closely log‐normal than the probability densities of the individual components. This is true regardless of whether &khgr; and its components are unaveraged or locally averaged over a linear dimensionr, in the Obukhov–Kolmogoroff sense. The variance &sgr;2of the logarithm of the locally averaged &khgr; is proportional to logrover a wide range ofr(rmax/rmin≃30), in contrast to the individual components where this ratio may be as small as 3. The value of the Kolmogoroff constant &mgr; determined from the slope of &sgr;2vs logris about 0.35. This is consistent with the slope of the spectral density of &khgr; and is also in agreement with previous best estimates of &mgr; obtained at high Reynolds numbers. Using only one component of &khgr;, the evaluation of &mgr; either from the slope of the spectral density or from the slope of &sgr;2vs logrseems to be highly ambiguous and can lead to erroneous results.