The objective of this study was to assess the interobserver reliability (agreement) and accuracy of keel palpation for the purpose of detecting old fractures in an end-of-lay flock of commercial laying hens. The low level of invasiveness and the relative speed at which this evaluation can be carried out lends itself well to use in a welfare audit, but only if the results are reliable and accurate from various assessors. The palpation technique first described by Wilkins and others (2004) was used to manually palpate for keel fractures. The technique was modified in that only keel fractures were considered. Eight assessors with varying laying hen experience palpated 100 live ISA Brown hens that had been in lay for 49 weeks. The hens were then euthanased and examined by dissection to establish whether there had been a keel fracture present (yes/no). The accuracy for individual assessors ranged from 87.1 to 96.8 per cent, with a mean of 91.8 per cent among all eight assessors. The interobserver reliability among all eight assessors was moderate (=0.44). Accuracy and κ values were 84.8 per cent and 0.41 for the first 50 hens, and 99.5 per cent and 0.47 for the last 50 hens, respectively, indicating that there was increased accuracy and agreement as the assessors became more experienced at palpation. This level of agreement, and the high level of accuracy, would make this technique an acceptable measure of keel fracture prevalence in a welfare audit.