Three computer-based experiments which tested human participants in a non-immersive virtual watermaze task sought to determine factors which dictate whether the presence of a visual platform disrupts locale learning and taxon learning. In Experiment 1, the visible platform disrupted locale but not taxon learning based on viewpoint-independent and viewpoint-dependent information, respectively. In Experiment 2, taxon learning based on non-geometric cues providing viewpoint-dependent information was disrupted by the visible platform when the cues required relational information to disambiguate them from other cues. Experiment 3 placed participants in an isosceles triangular pool. The presence of the visible platform did not disrupt the encoding of relational information provided by shape of the pool. These results support the notion that geometric cues are encoded in a separate module which is impenetrable to non-geometric cues not creating the shape of the environment.