Towards a fully predictive model of flight paths in pigeons navigating in the familiar area: prediction across differing individuals

There is strong evidence that pigeons use visual cues to navigate within the familiar area [1], and that visual landmarks constitute an important part of the familiar area map [2]. However, the precise role of the visual landscape in familiar area navigation remains controversial. In particular, the strongest evidence for visual landmark use – the development and recapitulation of idiosyncratic habitual homing routes – has been repeatedly observed in some locations [3, 4, 2], but not in others [5]. There is clear evidence that the specific nature of the release site influences whether pigeons will form familiar routes, and how idiosyncratic these may be [6]. The previous familiarity of birds with the landscape also effects whether the emergence of habitual routes is clearly observable (see e.g. [5]). Some release sites are associated with a clear set of possible ‘corridors’ for pigeons to learn, whereas others seem to lack this feature [2, 6] Some specific visual cues have been implicated in the homing process. Roads, rivers and other strong linear features appear to be particularly attrative to pigeons even on their first flights from unfamiliar sites, and habitual routes are often formed over these [3, 7]. Technical analysis of the landscape as presented in aerial photographs suggests that ‘edges’ (areas where the image changes abruptly) are associated with changes in navigational behaviour [8]. Recent work [9] has shown that pigeons’ fidelity to their previous routes is connected to the edge density of the underlying landscape, suggesting a link between visual complexity and navigational success, with optimal complexities found on the borders of forests and urban dwellings, as well as near roads and rivers. Visual inspection of candidate landmark sites suggest that other visually salient features such as church spires are likely to be used as visual cues as well [2]. All this evidence points to an emerging picture where certain types of landscape are more likely to be used as visual cues, leading to site-specific results when evaluating landscape use. The evidence so far suggests that strong linear features and the boundaries between different landscape types are particularly salient, and that pigeons have difficulty memorising landmarks in either visually barren (such as empty fields) or visually complex (such as inner urban or forest) landscapes. Nonetheless, because it is generally impossible to experimentally manipulate the landscape on a large scale, a complete understanding of how pigeons use visual cues has remained out of reach. It remains difficult to disentangle causality from correlation in these findings. However, while the landscape can not be manipulated directly, it is possible to choose new release sites to test specific hypotheses about the use of visual cues, if those hypotheses make sufficiently precise predictions about navigational behaviour. For this reason we propose to move towards a predictive model