Digital holography is a modern imaging technique whereby a propagated object wave interferes with a known (spherical or plane) reference wave at a plane where a digital sensor is situated. The resulting intensity distribution is recorded by a CCD or CMOS sensor array to produce a digital hologram. This digital hologram can be processed in several ways to isolate the real image term. Using a propagation algorithm, the object wave can be numerically reconstructed from this real image term. Several factors limit the performance of such imaging systems, such as the finite extent of the sensor array and the finite size of the equally spaced sensor pixels, which average the light intensity incident upon them. Theoretical results indicate that in a Fresnel-based system the role of these finite-size pixels is to attenuate higher spatial frequencies by convolving the reconstructed signal with a rectangular function of equal size to the light-sensitive area of the pixel. However, when a spherical reference wave is used, as is the case with "lensless" Fourier-based systems, spatial frequencies will not be attenuated; rather it is the complex amplitude of the reconstructed signal that will be attenuated. In this manuscript we explore this question in more detail, providing new theoretical and experimental results. By assuming a fully developed speckle field for the object wave, we examine the first-order statistical distributions for the integrated intensity of the object wave, and the interference term, using numerical simulations. We show that the statistical distribution of the interference term can be changed, by varying the sphericity of the reference wave. Experimental results are provided where we compare the performance of a Fresnel and Fourier holographic system as a function of pixel size.