Previous research has found a linear relation between distance and scanning times for spatial configurations when the spatial configurations were learned perceptually and when they were constructed from well-structured verbal descriptions. The current research replicated the time-distance relation when the images were constructed from repetitions of well-structured descriptions but not when the images were generated from three repetitions of a description that presented information in a random order. Six exposures to the randomly ordered information yielded the expected time-distance relation in image scanning. We posited that additional exposure to the poorly structured information allowed the image to develop the structural coherence and resolution needed to support consistent scanning. Thus, the structure of descriptions can affect the intrinsic structure of images of described objects and hence the mental operations performed subsequently on these images. Another experiment indicated that image coherence and resolution improves even after the verbal description is accurately recalled.