We critically re-examine the available data on the spectral types, masses a
nd radii of the secondary stars in cataclysmic variables (CVs) and low-mass
X-ray binaries (LMXBs), using the new catalogue of Ritter & Kolb as a star
ting point. We find there are 55 reliable spectral type determinations and
only 14 reliable mass determinations of CV secondary stars (10 and 5, respe
ctively, in the case of LMXBs). We derive new spectral type-period, mass-ra
dius, mass-period and radius-period relations, and compare them with theore
tical predictions. We find that CV secondary stars with orbital periods sho
rter than 7-8 h are, as a group, indistinguishable from main-sequence stars
in detached binaries. We find that it is not valid, however, to estimate t
he mass from the spectral type of the secondary star in CVs or LMXBs. We fi
nd that LMXB secondary stars show some evidence for evolution, with seconda
ry stars which are slightly too large for their mass. We show how the masse
s and radii of the secondary stars in CVs can be used to test the validity
of the disrupted magnetic braking model of CV evolution, but we find that t
he currently available data are not sufficiently accurate or numerous to al
low such an analysis. As well as considering secondary star masses, we also
discuss the masses of the white dwarfs in CVs, and find mean values of (M)
over bar(1) = 0.69 +/- 0.13 M-circle dot below the period gap, and (M) ove
r bar(1) = 0.80 +/- 0.22 M-circle dot above the period gap.