Numerous methods of predicting settlement of foundations on sands and gravels have been published-many more methods than for clays. The reason lies in the extreme difliculty of obtaining undisturbed samples for the laboratory determination of compressibility under appropriate conditions of stress and stress history. Hence resort has been made to the interpretation of field in situ tests such as the standard penetration test (SPT), cone penetration test and plate loading test, and much of the literature has been devoted to such interpretations. This extensive literature will not be reviewed here as it has been adequately covered by Sutherland,’Simons and Menzies’ and N i~ on.~ 2. The practical importance of the problem was perhaps put in perspective by Terzaghi4 when he stated that all buildings resting on sand which were known to him had settled less than 75 mm (3 in) whereas the settlement of buildings on clay foundations quite often exceeded 500mm (20 in). This statement provided the impetus for the study described in the present Paper in which a large number of case records of settlement on sands and gravels have been assembled by Burbidge’.

3. The essential details of most of these case records are tabulated in Appendix 1 of the present Paper and the associated references are given in Appendix 2. The case record numbering used by Burbidge has been retained for ease of reference. 4. The prime objective of the study was to check whether the above statement of Terzaghi’s still held true and reference to Appendix 1 shows that, with a few exceptions, it does for buildings. However, settlements well in excess of 75mm have been recorded for tanks and embankments on very loose sands. In view of the small settlements usually experienced with sands and gravels the second objective of the study was to analyse the data on actual observations of settlement employing a minimum of interpretation to see if a simple and useful picture emerged. A preliminary study of this type was undertaken by Burland et aL6 and a similar but more detailed approach is described here. 5. The picture that has emerged from the statistical analysis of over 200 cases is remarkably simple and gives a range of settlements which is generally less than the range of predictions offered by the current commonly accepted methods.’A brief description of the method and its application is given first, followed by a detailed account of the analysis of the settlement records.