Infrared spectroscopy has been used to provide an independent estimate of dipole orientation in poled films of a side-chain polymer based on poly(methyl methacrylate) with an oxynitrostilbene side group attached via an aliphatic spacer. In particular, we used the dichroism of the absorption band assigned to the NO2 symmetric stretching vibration as a measure of the orientation of the nitrostilbene group. This yields [P2(cos theta)], where theta is the angle between the transition dipole moment and the symmetry axis of the sample. We then used a Langevin-type model to calculate [cos theta] from [P2(cos theta)] and concluded that values of [cos theta] as high as 0.5 are being achieved. We find, however, that [P2(cos 0)] obtained by comparison of normal incidence measurements on unpoled and poled films does not agree with [P2(cos theta)] obtained from tilted-film measurements on poled films. We propose that this discrepancy arises because the local field in a poled sample affects the polarizability of the molecules. To confirm this, we applied high fields to unpoled films in the infrared spectrometer at room temperature. The results clearly show a large reduction in absorbance with the applied field when the infrared electric field vector has a component perpendicular to the plane of the film. A major error therefore occurs as a consequence of the internal field in poled samples if [P2(cos theta)] is derived from the tilted film measurements, whereas [P2(cos theta)] derived from normal incidence measurements is essentially correct.