Humans use a variety of strategies to reorient in space. There are diverging views on whether spatial reorientation relies on an encapsulated geometric module, an associative mechanism or an adaptive combination of different cues. We test these proposals with a computational model that predicts human behavior in reorientation. By analyzing existing data from multiple sources, we show evidence for an adaptive view of reorientation that combines information from geometry, polarization and language. Our work opens up opportunities to understand the interactive strategies of human reorientation.