Five non-destructive testing (n.d.t.) methods are widely used for defect detection: these are magnetic particle, dye penetrant, electrical eddy currents, radiography and ultrasonics. The first three can detect only surface-breaking or immediately sub-surface defects, while radiography and ultrasonics can also find embedded, remote defects. Ultrasonics is far more sensitive to cracks than is radiography; moreover, of all the n.d.t. methods, only ultrasonics can in general measure a crack's through-wall position and size. Consequently only ultrasonics is fully compatible with fracture mechanics requirements. Used in conjunction with fracture mechanics, ultrasonics has proved a powerful technique for demonstrating component integrity. After a brief description of the five main n.d.t. methods, the paper concentrates on ultrasonics. Basic ultrasonic techniques for detecting, positioning and sizing cracks are described and the main sources of error indicated. Two approaches to defect size assessment are recognized. The approach of endeavouring to measure crack size as accurately as possible is most appropriate to manual testing by skilled practitioners. The alternative involves measuring convenient parameters of the ultrasonic echo rather than of the defect itself, and aims to promote a speedy, reproducible examination. The subsequent discussion is of the implications that the limitations of n.d.t. techniques have for component design and fracture mechanics assessment. Among the points raised are the importance of access and component geometry, the need for cooperation in planning inspections and the ability of ultrasonics to distinguish significant from insignificant defects. The paper closes with two examples of the beneficial joint application of ultrasonics and fracture mechanics.