General plasma physics principles state that power flowQ(r) through a magnetic surface in a tokamak should scale asQ(r)= {32&pgr;2Rr3Te2c nea/[eB(a2−r2)2]}F(&rgr;*,&bgr;,&ngr;*,r/a,q,s,r/R,...) where the arguments ofFare local, nondimensional plasma parameters and nondimensional gradients. This paper reports an experimental determination of howFvaries with normalized gyroradius &rgr;*≡(2TeMi)1/2c/eBaand collisionality &ngr;*≡(R/r)3/2qR&ngr;e(me/ 2Te)1/2for discharges prepared so that other nondimensional parameters remain close to constant. Tokamak Fusion Test Reactor (TFTR) [D. M. Meadeetal., inPlasmaPhysicsandControlledNuclearFusionResearch, 1990, Proceedings of the 13th International Conference, Washington (International Atomic Energy Agency, Vienna, 1991), Vol. 1, p. 9] L‐mode data showFto be independent of &rgr;* and numerically small, corresponding to Bohm scaling with a small multiplicative constant. By contrast, most theories predict gyro‐Bohm scaling:F∝&rgr;*. Bohm scaling implies that the largest scale size for microinstability turbulence depends on machine size. Analysis of a collisionality scan finds Bohm‐normalized power flow to be independent of collisionality. Implications for future theory, experiment, and reactor extrapolations are discussed.