Binary stars are the main source of fundamental data on stellar masses and radii (M, R). Considerable progress has been made in recent years in the quality and quantity of such data, and stellar masses and radii of high accuracy have led to a number of qualitatively new and interesting results on the properties and evolution of normal stars. This paper reviews the current status of fundamental M and R determinations which (i) have errors less-than-or-equal-to 2%, the limit for non-trivial results in many applications, and (ii) can be presumed valid for single stars. These two conditions limit the discussion to data from detached, double-lined eclipsing binary systems. After a brief discussion (Sect. 2) of the main tests for accuracy and consistency which must be met for observational data to be included in the sample, data for 45 binary systems (90 single stars) are presented in Sect. 3 (Table 1 and Figs. 2-5). Spectral types are O8-M1 on the main sequence, with only two stars clearly in the red-giant region. From the review by Popper (1980), data for only 6 systems survive unchanged in the present list, while improved data are given for 18 systems; 21 systems are new additions. Broad-band colours, effective temperatures, and luminosities are also given, but are scale-dependent and considerably less reliably determined than M and R. The observed ranges in M and R for a given colour far exceed the observational errors, primarily due to evolutionary effects within the main sequence. For this reason, single-parameter relations used to predict M and R for single stars are limited to an accuracy of some +/- 15% in M and +/- 50% in R, basically independent of the number and accuracy of the data used to establish the relations. Two-parameter calibrations are discussed (Sect. 4) which can eventually reduce these errors to approximately 5% in both M and R. At this level, abundance effects become significant and presumably account for the residual scatter. Comparison of the data with steller evolution models is the topic of Sect. 5. Characteristic features of the data which are crucial in such work are emphasized, rather than attempts to "prove" the validity of any particular set of models. Already from M and R alone, some significant constraints can be derived (Fig. 4). When both M, R, and T(e) are known, the initial helium abundance Y can be estimated if the metal-abundance parameter Z is assumed or determined. Studies in which binaries with accurate values of M, R, and Z are fit by models calculated for the precise observed masses, and with Y and mixing length constrained to solar values, provide the most stringent tests of the models. Probing further model refinements such as convective overshooting requires full use of the potential of the data. For example, models may yield general main-sequence limits which are consistent with the observations, but still be unable to fit any single system to the precision of the data. Conditions for critical, informative tests are discussed. Tidal effects in binaries are briefly discussed in Sect. 6. As tidal forces are extremely sensitive to the dimensions and internal structure of the stars, the present sample is well suited for such studies. Recent success in matching computed and observed apsidal-motion parameters for early-type binaries is mentioned. Finally, main priorities for future work are outlined.
