Space syntax is generally geared towards identifying syntactic features in spatial layouts and predicting human usage patterns from configurations. From a psychological perspective it appears unlikely that people directly perceive such configurational features upon newly entering an environment and other perceptual features might have more direct impact on behaviour. In this study we experimentally juxtapose features typically captured by syntactic measures with perceptual features generally under-represented in such analyses, like local visual attractiveness of path choices. Research on indoor navigation (e.g. Haq & Zimring, 2003; Hölscher, Brösamle & Vrachliotis,2012) suggests that human route choices are reliably predicted by syntactic measures such as integration, both for free exploration of complex corridor layouts and for targeted wayfinding and search tasks. This paper reports a Virtual Reality wayfinding experiment that tests the relative impact of syntactic properties and perceptual attractiveness. For this purpose we developed an experimental indoor environment that represents a roughly L-shaped hospital building with two main corridors passing through the middle. Syntactically, these two corridors had the highest integration (HH) and connectivity values of all the corridors in the building. These corridors have a total of 22 intersections and form the syntactic integration core (Peponis, Zimring & Choi, 1990) of the building. Two versions of this layout were programmed in a CAD tool: in version A (the control or baseline condition) all corridors were of the same colour, width and wall texture. In version B (the experimental condition) a sequence of corridor segments was visually highlighted by a bright colour, different textures and an increased width between walls. This path was of equal metric length as the syntactic integration core but contained fewer, only 9, intersections. The visual features were selected so that they were salient and provided characteristics that are typically associated with important main corridors in a building. In the experimental condition B, this highlighted corridor sequence can be considered a “fake integration core”. The main hypothesis for the study is that this fake integration core would be used more in the visually highlighted condition and thus take activity away from the syntactic core and reduce the predictive power of syntactic features. Layout variation was implemented as a between-participants factor, i.e. each of the 42 participants experienced only one version of the building. They explored the layout for five minutes, and were then taken along prescribed routes through the building to learn the location of four landmarks. Navigation to these landmarks was tested as well as search time for two additional new locations. Analysis of path choices reveals a complex pattern of environmental and individual influences on movement patterns, including the location of start points and landmarks as well as idiosyncratic preferences. While the syntactic integration core has a substantial level of usage, highlighting the fake integration core strongly increases local movement attracted by local perceptual features. Overall this psychological experiment suggests that syntactic analysis and perceptual properties of an environment need to be considered to appropriately capture human movement behaviour.