Understanding the molecular and cellular mechanisms that mediate magnetosensation in vertebrates is a formidable scientific problem(1,2). One hypothesis is that magnetic information is transduced into neuronal impulses by using a magnetite-based magnetoreceptor(3,4). Previous studies claim to have identified a magnetic sense system in the pigeon, common to avian species, which consists of magnetite-containing trigeminal afferents located at six specific loci in the rostral subepidermis of the beak(5-8). These studies have been widely accepted in the field and heavily relied upon by both behavioural biologists and physicists(9-11). Here we show that clusters of iron-rich cells in the rostro-medial upper beak of the pigeon Columbia livia are macrophages, not magnetosensitive neurons. Our systematic characterization of the pigeon upper beak identified iron-rich cells in the stratum laxum of the subepidermis, the basal region of the respiratory epithelium and the apex of feather follicles. Using a three-dimensional blueprint of the pigeon beak created by magnetic resonance imaging and computed tomography, we mapped the location of iron-rich cells, revealing unexpected variation in their distribution and number-an observation that is inconsistent with a role in magnetic sensation. Ultrastructure analysis of these cells, which are not unique to the beak, showed that their subcellular architecture includes ferritin-like granules, siderosomes, haemosiderin and filopodia, characteristics of iron-rich macrophages. Our conclusion that these cells are macrophages and not magnetosensitive neurons is supported by immunohistological studies showing co-localization with the antigen-presenting molecule major histocompatibility complex class II. Our work necessitates a renewed search for the true magnetite-dependent magnetoreceptor in birds.