The fatigue initiation behaviour of two Al-based bearing lining alloys has been assessed by image analysis tessellation techniques and subsequent adaptive numerical classification approaches. Fatigue initiation occurs by decohesion at Si particles, and in the absence of Si particles decohesion occurs at Sn particles. In both cases, relatively unclustered, large particles initiate fatigue, and the relative position of the nearest neighbour particle seems to have significance. This can be explained qualitatively in terms of strain mismatch arising from differing elastic moduli between the particles and surrounding Al matrix and local stress concentration and superposition effects. The lining alloy containing no Si and high Sn content exhibits more directionality, and a heavier dependence on angular information is found by the classifier.