ABSTRACT Large scale specimens of X-joints were tested in fatigue, either in tension or in in-plane bending (R=O.1). Similar specimens (size ratio 1/2) were tested in order to estimate the scale effect. y was chosen here from 8 to 12, at the limit of the validity of any shell-theory. In each node, one brace joint was heat-treated and the other was left as welded. The specimens were instrumented by strain gauges. During monotonic static loading, the measured distribution of stresses shows that the Effective stress concentration factor KE can be defined as a product of two factors KE =KG × KL, where KG reflects the Global geometry effect of the intersection of the two shells and KL the influence of the Local shape at the weld toe. During the fatigue test, the crack growth and the evolution of strains along the weld-toe were monitored. On the basis of the obtained results an elementary theoretical model for life prediction is proposed which takes into account the stress redistribution due to the crack growth. INTRODUCTION The fatigue analysis of structures in general and offshore structures in particular is based on the comparison of a design stress with a specified S-N curve. This curve is usually obtained as the most conservative envelope of a constellation of dispersed points, corresponding to fatigue results obtained on different small-size and large-size models, put in a more or less clearly defined S-N coordinate system. The recent progress in fracture mechanics should allow a more rational approach to the definition both of the design stress (S) and of the fatigue life (N). As a part of a research sponsored by the E CS C (European Coal and steel Community) the IRSID (French Steelmaker Research Institute) has tested ten X joints. In parallel, a French programme of theoretical and numerical studies was conducted at the Ecole Polytechnique in order to give a basis for the interpretation of test results, which would ensure more reliable design data for engineers. The present paper gives an account of a part of this research. After a brief survey of the test programme, an approach to the definition of the design stress is presented in § 3 and an elementary model of the fatigue crack growth in tubular joints is proposed in § 5, which may help a more rational discussion of S-N curves. TESTING DEVICES Test specimens The test specimens consisted of large size welded X-joints, represented in Fig. 1. The average values of actual dimensions of chords and braces, in particular the tube thicknesses, as well as the values of the usual non-dimensional ratios and the types of loading are given in Table 1. The specimens were made from grade E 36 steel plate, according to AF NOR Standard. They were fabricated by a manufacturer familiar with offshore structures according to a specification which is typical of that currently used in North Sea practice. Weld size and configuration were chosen in accordance with API codes.