Weak lensing by large-scale mass inhomogeneities in the Universe induces co
rrelations in the observed ellipticities of distant sources. We first revie
w the harmonic analysis and statistics required of these correlations and d
iscuss calculations for the predicted signal. We consider the ellipticity c
orrelation function, the mean-square ellipticity, the ellipticity power spe
ctrum and a global maximum-likelihood analysis to isolate a weak-lensing si
gnal from the data. Estimates for the sensitivity of a survey of a given ar
ea, surface density, and mean intrinsic source ellipticity are presented. W
e then apply our results to the FIRST radio-source survey. We predict an rm
s ellipticity of roughly 0.011 in 1 x 1 deg(2) pixels and 0.018 in 20 x 20
arcmin(2) pixels if the power spectrum is normalized to sigma(8)Omega(0.53)
= 0.6, as indicated by the cluster abundance. The signal is significantly
larger in some models if the power spectrum is normalized instead to the CO
BE anisotropy. The uncertainty in the predictions from imprecise knowledge
of the FIRST redshift distribution is about 25 per cent in the rms elliptic
ity. We show that FIRST should be able to make a statistically significant
detection of a weak-lensing signal for cluster-abundance-normalized power s
pectra.