Materials with low coefficients of linear expansion are of considerable interest. The following are the basic specifications for dilatometers to be used with them: stability in the displacement sensor and small contributions from the components to the measured thermal expansion. Only interference dilatometers meet those requirements strictly, but new dilatometers have been described [i] based on the measurement of electrical quantities whose sensitivity is greater than that of an interference dilatometer, while not completely meeting those requirements. Here we describe a capacitance dilatometer for the range 300-5 K, which combines high stability in the displacement sensor with almost complete absence of thermal expansion in the body. The means of excluding the thermal expansion in the body is derived from [2], but we used a radio-frequency oscillator because of the lack of good Soviet capacitance bridges. Also, to improve specimen installation for routine measurements, the temperature monitoring system was mounted in a special unit, which was in thermal contact with the specimen. The dilatometer cell (Fig. I) is novel. The capacitance change is determined from the frequency of the oscillator present in the cryostat along with the cell. The cryostat is not subject to any particular specifications, as it can be a circulation one or have a system of vessels containing coolants linked to the cell by a cold finger. The cell and oscillator are linked by a common copper body; copper provides good thermostatic control (!3 x 10 .4 K), and thus stability in size during the measurement in spite of the high temperature coefficient of expansion for copper. The specimen 1 is heated independently of the body 2, which is provided by the support 3 with specimen resting on the insulators 4. The measurement thermometer 5 and control thermometer 6 work with the heater 7 to provide a given temperature in the support. Thermal contact between the support and the specimen occurs over a fairly large area, and in addition, the specimen is enclosed in the foil 9, which also has thermal contact with the support. We used cylinders i00 0.i mm long and 15-25 mm in diameter with ground flat ends. The large diameter was chosen to give better temperature equalization for materials with low thermal conductivity. A difference from [2] was that the support enabled one to mount the specimen on the thermal insulators together with the heating and thermostatic control system and simplified the specimen mounting. The support should not participate in the thermal expansion, which was a task handled as in [3]. The support consisted of two mating copper parts held by the bolts 8. The plane of contact between the upper disk is the plane of support containing the insulators, while the plane of contact for the lower part is the plane of contact with the specimen. When the temperature varies, the position of the plane of contact in space does not alter, so the expansion of the substrate does not influence the measured expansion of the specimen.