An important parameter for observational studies on dynamical variations of the ocean is temperature (T) which determines density variations together with salinity. While the achievement of high absolute accuracy of T-sensors is a formidable task, even the maintaining of high precision is difficult for any long-term deep-ocean observations. This is because the deep ocean is very weakly stably stratified in density, in which all the dynamics is covered by a range of about 0.001 °C. One requires high-resolution T-sensors with low noise levels and precisely calibrated. However, common Negative Temperature Coefficient (NTC)thermistor T-sensors have electronic drift causing a bias of temperature with time. The drift rate varies with temperature (change) and time (age) between 0.0002 and 0.001 °C wk−1 for T-sensors used here. A string of such T-sensors moored in the open-ocean for more than a month after calibration requires a drift-correction. When moored in the deep-ocean, a secondary correction is necessary even immediately after calibration. Secondary correction uses reference layers that are homogeneous to within 0.00005 °C over 100 m vertically plus an additional polynomial correction to remove remaining bias prior or after the reference date. Secondary correction may also involve pressure information on surface tides and mooring deflection. In this paper, secondary corrections are described for two cases of data. For one case from the West-Mediterranean varying temperature-density relationship and near-homogeneous conditions exist. The other is from hadal-deep instruments on a long mooring line in the Challenger Deep, Mariana Trench, affected by tidal flow. Nearby shipborne CTD-data are always needed to establish the temperature-density relationship and the background vertical temperature slope to which the moored T-sensor data are referenced. The stability of moored T-sensors is compared with that of temperature sensors of CTD.