TIGER integral-field spectrography and Hubble Space Telescope WFPC2 imaging of the E3 galaxy NGC 2974 are used to derive the kinematics of the stellar and ionized gas components in its central 500 pc . We derive a numerical two-integral distribution function from a Multi-Gaussian Expansion ( MGE ) mass model using the Hunter & Qian formalism . The TIGER as well as published long-slit stellar kinematics , including higher order moments , are well fitted with this self-consistent model , requiring neither the addition of a significant mass contribution from a hidden disc structure , nor the presence of a central dark mass ( at that spatial resolution ) . The data reveal the presence of a striking , highly contrasted , two-arm gaseous spiral structure within a radius of \sim 200 pc , corresponding to a total mass of 6.8 \times 10 ^ { 4 } solar masses of ionized gas . We use a deconvolved TIGER datacube to probe its kinematics at a resolution of about 0 \aas@@fstack { \prime \prime } 35 FWHM . Strong departures from circular motions are observed , as well as high velocity dispersion values on the inner side of the arms . We interpret the observed gas morphology and kinematics as the signature of streaming gas flows driven by a \sim 540 pc diameter bar with \Omega _ { p } = 700 \pm 100 \mbox { km s } ^ { -1 } kpc ^ { -1 } . This hypothesis is strongly supported by the predictions of a density wave model . This model predicts that the bar should lie at about 35° from the line of nodes , and implies gas inflow towards the central \sim 50 pc . The quadrupole pertubation due to this bar is estimated to represent less than 2 % of the underlying gravitational potential ( a maximum torque of about 10 % ) , explaining the lack of a direct detection via broad-band photometry in the visible . Despite its weakness , the inner bar of NGC 2974 may be able to drive some gas within a 10 pc radius . We suggest that the presence of such inner bars might be more common among early-type disk galaxies than is generally thought , and that deep high-resolution emission-line imagery may be the best way to detect such structures .