The use of a fracture mechanics test to evaluate the joint strength through the determination of the strain energy release rate G is nowadays well established. The joint strength for fluorinated polymer (PVDF) sheets bonded with an epoxy adhesive was studied using a double cantilever beam (DCB). In order to obtain small-scale yielding, the adhesive joint of the polymer specimens was strengthened by steel sheets. Pre-cracks were initiated at the center of the bond thickness separating the two PVDF surfaces, with nominal lengths ranging from 5 to 27.5 mm. We did not measure the evolution of the crack length, which is generally very difficult to obtain with good precision. The measurement of the load-point displacement was used instead. The opening load versus this load-point displacement was recorded. The slope of the first part of this curve gives the value of the initial stiffness of the joint specimen. The stiffness of the various specimens enables us to access the real experimental initial crack length, which was smaller than the nominal value, by comparison of the experimental values with the numerical ones. From the second part of the curve, the strain energy release rate values for the crack propagation in the initial step (G(1)) and in the steady step (G,) are deduced. They were calculated from a least-squares linear fit obtained from the load-point displacement versus the inverse square of the load curve. The experimental results are discussed in light of an analytical analysis using the thin beams approach, improved with an elastic foundation model developed by Maugis describing the deformation of materials behind the crack tip, and of a numerical approach based on a finite element analysis. In this numerical model, an elastic-plastic behavior of the materials has been assumed. Analytical and numerical approaches are compared and their validity and limitations are discussed.