When a scene changes between viewing instances, humans make relatively more fixations to, dwell longer on, and make more transitions in and out of the altered object’s area. This is typically shown as differences between novel, repeated, and manipulated scenes. Here, we explored whether not only the existence but also the extent of learning history prior the scene change affected viewing behavior. Experiment 1 adapted a paradigm by Ryan et al. (2000) to establish the memory effects with computer-generated scenes and high-resolution eye-tracking. Subjects viewed a set of scenes for a later memory test. Scenes were repeated, manipulated (object shifting or addition/deletion), or novel. Implicit relational memory became visible for shifting in all and for additions in most measures. In Experiment 2, we varied the extent of learning by repeating the scenes 1, 2, 4, or 8 times before introducing the manipulation. Based on the first experiment’s results, all scenes underwent a shifting manipulation. We calculated the difference in the critical object’s viewing parameters between the presentation before the change and the change presentation. The number of repetitions had a significant effect. Participants increased their dwell time, fixation proportion, and number of transitions more when they had viewed the scene for 4 or 8 times than for 1 or 2 times. Eye movement measures correlated with participants’ explicit memory between .37 and .51, yet not significantly. Thus, the investigated eye movement measures may capture memory strength rather than memory presence alone, but their suitability for memory strength assessment is discussed.