The potential sources of systematic uncertainty related to studying electron temperatures are examined in the context of deriving element abundances in H II regions. An illustration of the problem is developed for forbidden S III in Galactic H II regions, and attention is given to the overestimation of electron temperature T(e) due to the thermal structures of metal-rich nebulae. Accurate abundances for ions such as S(+2) and Ar(+2) are shown to require good characterizations of the temperature of each ion within a nebula, whereas some groups of ions can be described by common values of T(e). In order to measure abundances from collisionally excited lines in the UV it is important to avoid using lines which are sensitive to T(e) and to employ emission-line ratios with disparate temperature dependencies. The thermal gradient in metal-rich nebulae can lead to forbidden-line diagnostic ratios that describe electron temperatures far larger than the ion-weighted average values.