Atom column intensities of silicon single crystals oriented along different crystallographic orientations are compared in experimental and simulated high-angle annular dark-field images in scanning transmission electron microscopy. The intensity of a background, measured between the columns, is also evaluated and found to be largely independent of the column spacing. The contrast is lower in the experiments, and it has been suggested previously that this follows because the background is higher in the experiments. We explore the extent to which the comparison of experimental atom column intensities with image simulations is aided by subtraction of the background for these data. We also explore an alternative view: simple simulations overestimate the contrast because spatial incoherence and associated instabilities are not taken into account. The results do not distinguish between these approaches, which need not be in opposition, and we describe experiments which might further clarify matters.