Résumé
We rewiew and scrutinize the existing mass determinations of the pentaquarks from the exponential Laplace Sum Rules (LSR). We do not find any sum rule window for extracting optimal and reliable results from the LSR, due to the unusual slow convergence of the OPE and to the exceptional important role of the QCD continuum into the spectral function in this channel. Instead, we use in this channel,for the first time, Finite Energy Sum Rules (FESR), which exhibit a nice stability in the QCD continuum threshold t_c, at which one can extract, with a good accuracy, the mass of the lowest resonance. Including the D=7, 9 condensate contributions in the OPE, we obtain M_Theta=(1513+- 114) MeV, and the corresponding residue lambda_Theta^2= -(0.14-- 0.49)x 10^{-9} GeV^{12}, which favours the I=0, J=1/2, and negative parity S-wave interpretation of the Theta (1540). However, our analysis indicates a degeneracy between the unmixed I=0 and I=1 S-wave states. In the I=0, J=1/2, P-wave channel, we obtain, for the P-resonance, M_P = (1.99+- 0.19) GeV and lambda_P= -(0.7--7.1)x 10^{-9} GeV^{14}, which we expect to be discovered experimentally. Our results also suggest that some intuitive choices of the continuum threshold used in the LSR literature are inconsistent with the FESR results. Finally, a study of the Theta-K-N coupling using a vertex sum rule shows that, for the I=0, S-wave channel, the leading OPE contributions only start to order alpha_s^2 in the chiral limit m_s=0, indicating that the Theta is very narrow.