Abstract
Our understanding of bird orientation guided by magnetic and visual cues is primarily based on Emlen funnel experiments. Migration-motivated birds jump in the direction they want to fly, and their feet leave marks on paper lining the funnel, which yields the preferred direction. Despite the low signal-to-noise ratio, this paradigm has proven instrumental for studying magnetoreception in birds. However, the high noise limits the questions that can be answered and there is no data-informed guideline for selecting sample sizes that have a high likelihood to be conclusive. Furthermore, differences in experimental design traditions limit comparison and reproducibility across studies, slowing down discovery. We performed a large meta-analysis across double-blind magnetic orientation studies with Emlen funnels performed at Oldenburg to statistically characterize Emlen funnel data and determine minimal sampling requirements for conclusive experimental design. The analysis confirms that pre-selecting migration-motivated animals before the real experiments start improves statistical power by reducing noise. We also highlight mathematical limitations of the widely used directionality measure 'r', due to lacking sample-size bias correction, and present realistic ranges for expected bird directedness in Emlen funnels. Combined, these results provide critical design and analysis guidelines for statistically informative magnetic orientation experiments.