As the shipbuilding industry has emerged from an extended recession, orders for high-value-added ships, such as LNG and ultra-large container ships, are increasing. For ultra-large container ships, high-strength, thick materials are applied. Because the possibility of brittle fracture increases owing to the application of thick steel plates, the related regulations of the International Association of Classification Societies have been strengthened to prevent brittle fracture. To secure brittle fracture stability, it is necessary to secure crack arrest toughness (Kca) through large ESSO experiments or to secure a crack arrest temperature (CAT) value. Because large-scale experiments require considerable costs and efforts, efforts have increased to examine brittle fracture stability through small-scale tests. In the present study, a technology was developed to predict CAT with small specimens. The CAT prediction formula developed with small specimens makes it possible to accurately predict CAT using data obtained through large-scale experiments.