Independent temperature microstructure samples of Gregg (1977) in the main thermocline of the mid‐Pacific in different years and seasons are used to infer space‐time average vertical heat fluxes and space‐time average turbulence and temperature dissipation rates. Lognormality plots of these data suggest that mixing rates in thermocline layers are approximately lognormal and very intermittent, with intermittency factors σ2lnC̄150m that increase with depth from 3 to 6 as Kz increases from ≈ 0.2 to 1.6 cm2 s−11, and ϵ ≈ (0.6–1.6)×10−4 cm2 s−3. Decorrelation length scale ranges for turbulence with such large intermittency factors should be large by the Kolmogorov intermittency (third universal similarity) hypothesis. Available microstructure data suggest that mixing concentrates near fronts and is dominated by powerful bursts of turbulence that fossilize and radiate breaking internal waves. Equations for the estimation of vertical heat flux from temperature dissipation rate measurements are rederived by a control volume rather than Reynolds averaging method that emphasizes the necessity for large averaging length and time scales and does not exclude frontal mixing and nonturbulent mixing processes that are neglected by most applications of the Osborn and Cox (1972) model. The indicated heat fluxes qz are about 6 and 3 W m−2 at 0.1 and 1 km depths, consistent with deeper bulk flow estimates of qz ≈ 2 W m−2 from Munk (1966) but much larger than those inferred from the Gregg (1989) thermocline ε correlation that decrease from qz ≈ 1 to 0.1 W m−2 in this depth range. It is suggested that the Gregg (1989) values of qz, ̄2lnϵ ≈ 0.04–0.9, Kz ≈ 0.06 cm2s−1, and ϵ ≈ 7×10−6 cm2 s−3 for the ocean main thermocline are underestimates due to a combination of undersampling errors.