Lens-source cross-correlations as a test of source galaxy position deflections. We show two different lens-source ɀ-bin pairs for galaxy samples cut at I_E < 26: a case with ɀ_d ≃ 1 and ɀ_s ≃ 2 in the top panel, ɀ_d ≃ 1.5 and ɀ_s ≃ 3 in the bottom panel, respectively. Flagship measurements (filled blue circles) are compared to linear and nonlinear theory predictions (lines). Lower panels show the ratio of FS2 measurements over nonlinear theory prediction. Sample variance errors around nonlinear theory are displayed as dotted lines. Cross-correlation in absence of magnification (i.e, noise estimate) is shown as open red circles. The FS2 measurements use an estimator that roughly cancels sample-variance (see text for details).