The polar format algorithm (PFA) allows the use of computationally efficient fast Fourier transforms in synthetic aperture radar (SAR) image formation, but introduces phase errors when making the far-field approximations that facilitate this approach. The phase errors cause spatially variant distortion and defocus in the formed image. These effects may complicate target recognition applications. To limit the impact of defocus, scene size is usually limited such that the maximum quadratic phase error within an image falls below some threshold. This paper looks at how distortion and defocus affects the classification of targets, with the hope of developing an application-driven scene size limit.