On the basis of detailed considerations of the relevant effects and parameters, we describe techniques which have been developed to accurately set, maintain, and measure the temperature of a small sample (typically a single‐crystal disk of about 1 cm in diameter and 1 mm thick) in a strongly nonisothermal surrounding in ultrahigh vacuum, such as required by surface science experiments, in the temperature range from at least 6.5 K up. The resettability and resolution are 30–70 mK and the absolute error is estimated as about 0.7 K at 10 K and about 1.5 K at 100 K. Controlled linear heating with rates from 10−2to 50 K/s and high constancy (deviations below 0.1 K up to 5 K/s and below 0.5 K up to 50 K/s), and stepwise heating (≳100 K/s) without measurable overshoot can be carried out in this whole cryogenic range.