The analysis of the shear induced by a single cluster on the images of a large number of background galaxies is all centered around the curl-free character of a well-known vector field that can be derived from the data. Such basic property breaks down when the source galaxies happen to be observed through two clusters at different redshifts, partially aligned along the line of sight. In this paper we address the study of double lenses and obtain five main results. (i) First we generalize the procedure to extract the available information, contained in the observed shear field, from the case of a single lens to that of a double lens. (ii) Then we evaluate the possibility of detecting the signature of double lensing given the known properties of the distribution of clusters of galaxies. (iii) As a different astrophysical application, we demonstrate how the method can be used to detect the presence of a dark cluster that might happen to be partially aligned with a bright cluster studied in terms of statistical lensing. (iv) In addition, we show that the redshift distribution of the source galaxies, which in principle might also contribute to break the curl-free character of the shear field, actually produces systematic effects typically two orders of magnitude smaller than the double lensing effects we are focusing on. (v) Remarkably, a discussion of relevant contributions to the noise of the shear measurement has brought up an intrinsic limitation of weak lensing analyses, since one specific contribution, associated with the presence of a non-vanishing two-galaxy correlation function, turns out not to decrease with the density of source galaxies (and thus with the depth of the observations).

40 pages, 15 figures. Accepted for publication in ApJ main journal