Presented at the 2014 Ocean Sciences MeetingIt has long been recognized that ocean salinity estimates made from in situ conductivity-temperature-depth (CTD) instruments depend for accuracy on correction of sample temperature for thermal exchange across conductivity cells. Most commonly, a filter predicated on unit step response modeled by a single exponential of arbitrary amplitude and decay time scale has is used, where scale parameters are chosen empirically. Radial heat exchange through a cell including flux across both the inner glass and exterior jacket walls can be expressed as a sum of cylindrical thermal normal modes. Summing over the first few modes adequately models the complete solution. For typical flow rates past standard Seabird Electronics conductivity cells, roughly half the heat flow across the interior glass wall acts to alter temperature of the surrounding jacket. The empirical single exponential model over-corrects at short time scales and under-corrects at longer time scales, leading to spurious salinity signals and induction of artificial variance in temperature-salinity space. The heat exchange model applied to arbitrary flow rates both within and outside the cell demonstrates that thermal inertia correction for un-pumped cells on underwater gliders is no greater than for pumped cells on ship-lowered CTDs. Intermittent pumping as used on some floats induces a salinity bias comparable to what has been attributed to climate change.