ABSTRACT A data base containing 750 the ultimate strength test results on" of grouted connections has been assessed in order to provide improved design guidance. The test data have been screened in order to obtain an assessment of available design equations. The data have also been systematically studied to identify new trends discernible from this large data base which has not previously been available as an entity. Supplementary tests have been performed on specimens with high h/s ratios and to obtain data on grout to steel interface friction. The screened test data have finally been used in a calibration study on the design formulations used in three different codes. - API RP 2A, DNV and D. En. (U. K.). DNV's structural reliability programme "PROBAN" has been used to estimate safety factors, load factors and material factors required to achieve the same target reliability as aimed for in the new API RP 2A LRFD code. Significant differences between the safety levels in the D. En. Guidance and those entrenched in the API and the DNV codes are indicated. The study indicates that the reliability levels entrenched in the API and DNV codes are in line with those adopted in structural design in general. Sensitivity studies have been performed to support the reliability calculations. Variations in the design equations as well as the adopted safety margins give rise to very large differences in required connection lengths. The more conservative requirements may lead to as much as a 200 % increase in connection lengths. FAILURE MODES OF GROUTED CONNECTIONS The failure modes observed to occur in grouted connections depend on whether shear keys are present and their size and spacing: PLAIN GROUTED CONNECTIONS have no shear keys. In this case failure occurs by slipping along the interface between the pile and the grout. On dissection some tensile cracks may be observed in the grout. These are, however, of a secondary nature and principally due to shrinkage and not relevant to the failure. All relative displacement takes place at the pile surface. The connection behaves as two rigid elastic bodies. It can be observed that tensile hoop stresses are set up in the sleeve and compressive hoop stresses in the pile. This due to the wedging action caused by the uneven surface of rolled steel tubulars. Tests performed by DNV using tubulars which have been turned down in a lathe to produce a uniform surface show a radical reduction in load carrying capacity. % strength of plain pile connections is, therefore, dependant on the magnitude of the wall surface unevenness as well as on the hoop stiff nesses of the pile and sleeve. This can be expressed by the sum of the thickness to diameter ratios of the tubulars and will determine the radial stress set up by the wedging action and, thus, the inteface friction force that will be mobilized. This is developed in detail in reference /1/.