In computing ionic abundances in H II regions, a two-zone model for the electron temperature, T(e), is often assumed, with one temperature assigned to a "low-ionization" zone for species such as O+ and N+, and another temperature assigned to a "high-ionization" zone for species such as O+2 and Ne+2. Photoionization models, however, suggest that some ions, such as S+2 and Ar+2, do not fit easily into such a scheme; in such cases, an intermediate value for T(e) is more appropriate. It is shown that this is an important consideration when computing S+2 abundances from measurements of the [S III] lambda-6312 line, which has a strong temperature dependence. Knowledge of the proper prescription is necessary for determining whether the Shaver et al. [MNRAS, 204, 53 (1983)] measurements of [S III] in Galactic H II regions yield a gradient in S/O across the Galactic disk. Examination of the behavior of T(e) in the photoionization models suggests that T(e) for S+2 and Ar+2 follows the relation T(Ar+2) = T(S+2) = 0.83 T(O+2) + 1700 K. Meanwhile, the models suggest that T(N+) = T(O+) = T(S+) and T(N+2) = T(O+2) = T(Ne+2)= T(C+2) are reasonably good approximations. In any case, when estimating abundances it is best to avoid using lines which are extremely sensitive to T(e), or to use ratios of emission lines which have similar temperature dependencies, if possible. This strategy will be important for measuring abundances from collisionally excited lines in the ultraviolet spectrum. In a cool, metal-rich H II region, radiation hardening, combined with strong fine structure line cooling in the interior of the nebula, leads to a thermal structure in which the electron temperature increases radially outward, as shown previously by Stasinska. This thermal gradient causes the forbidden line diagnostic ratios to yield electron temperatures which can be significantly larger than the ion-weighted average values; photoionization models indicate that the effect becomes important for T(e) < 8000 K for [N II] and T(e) < 10 000 K for [O II]. The result is that measured electron temperatures can lead to too small derived abundances in such nebulae.