© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement In several modes of retinal imaging, the primary means of visualizing cone photoreceptors is from reflected light. Understanding how such images are formed, particularly when adaptive optics techniques are used, will help to guide their interpretation. Toward this end, we used finite difference beam propagation to model reflections from cone photoreceptors. We investigated the formation of cone images in adaptive optics scanning laser ophthalmoscopy (AOSLO) and optical coherence tomography (AOOCT). Three cone models were tested, one made up of three segments of varying refractive index, the other two having additional boundaries at the inner/outer segment junction and outer segment tip. Images formed by the first model did not correspond to AOOCT observations in the literature, while the latter two did. The predicted distributions of reflected light intensity from the latter cone models were compared to the distribution from AOSLO images, both studied with light sources of varied coherence length. The cone model with the most reflections at the inner/outer segment junction best fit the data measured in vivo. These results show that variance in cone reflection can originate from light interfering from reflectors much more closely spaced than the outer segment length. We also show that subtracting images taken with different coherence length sources highlights these changes in interference. Differential coherence images of cones occasionally revealed an annular reflection profile, which modeling showed to be very sensitive to cone size and the gaps bracketing the outer segment, suggesting that such imaging may be useful for probing photoreceptor morphology.