To evaluate the performance with respect to selectivity of the effect of the wings bending in the cone coil relative to the double coil in transcranial magnetic stimulation. The focal area and the width vector of the central lobe of the induced electrical field distributed along an elliptic surface approximating the cortex were computed for four coil models. The models represented the real coils, the double B70 and the cone B80 Medtronic, and their corresponding simulated flat coils (B70flat and B80flat). A response function was evaluated in 10 subjects for distal and proximal muscles of the upper limb by stimulation of the motor cortex along a line approximating the central sulcus. The width of the response function, at the level of the center of gravity, provided a quantitative measure for coil focality. The focal area for B70, B70flat, B80, and B80flat calculated from the model was 31.4, 32.2, 94.4, and 50.6 cm2, respectively. The width of the central lobe along the stimulation line was: 36.2, 37, 46, and 48.6 mm, respectively. Mean values of focality measure obtained experimentally were in distal muscles, 5.06 RPU (relative position units) for B70 and 5.99 RPU for B80; in proximal muscles, 4.11 RPU for B70 and 5.13 RPU for B80, with a mean RPU value of 11.13 mm. The difference, a 19% focality measure increase in B80 relative to B70 in distal muscles, was statistically significant (P < 0.001). The focality was demonstrated to be highest for the double coil. The width of the central lobe of the induced electrical field distribution is well reflected in the width of the response function. The increase in B80 is mainly due to wing geometry and relative placement of wings and is not due to the wing bending. The width of the central lobe characterizes the spread of the induced current below the wing junction, and it is a better focality estimator than the focal area for cone coils.