Consider a ferromagnetic crystal at small fields. In a domain of type i (i = 1, 2,…,n), the spontaneous magnetization lies along the ith direction of easy magnetization; the free-energy density associated with externally controlled fields is Ei, and that associated with random internal forces is εi, Postulate: at each point, ε1, ε2, …, εn are independent random variables, with a common distribution function F(x);, and the actual orientation is that of smallest Ei+εi. Let pj be the probability that a randomly chosen point is in a domain of type j. Then pj = ∫−∞∞{Πi≠j[1−F(x−Ei)]}F′(x−Ej)dx. When E1 = E2 = … = 0,the integrand is 1/n times the probability density Ψn′(x)of the smallest value among the n εi's. Under certain conditions,the distribution function Ψn(x) of the smallest value reduces asymptotically, for large n, to 1−exp[−eαn(x−un)], where F(un) = 1/n and αn = nF′(un). Under the same conditions, pi reduces asymptotically to e−αnEj/Σie−αnEi. This Boltzmann-type formula,variously “derived,” was used in an old “statistical” domain theory recently revived. Its limited success is understandable when it is interpreted as an asymptotic approximation, since n is usually only 2, 6, or 8. Use of the rigorous formula, with a reasonable form for F(x), might give better results.