One approach to the prediction of notch fatigue limits is the Neuber method in which stresses are averaged over a critical distance ahead of the notch. In recent years this theory has been updated by the discovery of an analytical method for finding the critical distance for a given material, which shows that the appropriate distance is 2ao , ao being El Haddad’s short‐crack constant. The present author has advocated this approach, which he has called the Line Method (LM) and has tested it extensively against experimental data. However the method still remains essentially empirical; the aim of the present paper was to try to link this approach to the known mechanisms of crack growth at a notch. It is proposed that the LM is successful because it expresses the conditions necessary for growth of a small crack located at the notch root. The argument is developed by starting from the resistance curves used to predict non‐propagating cracks and by linking the LM with the expression for stress intensity, K, derived by the crack‐line loading method. The results provide some mechanistic justification for the use of the LM for sharp, crack‐like notches; its success for other types of notch (i.e. blunt notches and short notches) requires a different explanation.