Engineers Australia

On the common criticism of initial moisture content bias in the shrink-swell test

Stephen Fityus and Glen Burton


The shrink-swell test, and its application in site classification and foundation design according to AS2870 (2011) has both advocates and critics. It has been used as the basis for many residential foundation designs in Australia for more than 30 years, and although examples of performance failures in residential structures are familiar to most practitioners, the incidence is relatively low compared with the number of structures that are constructed, and substantiated examples of widespread and systematic failures due to poor foundation design are few and far between. Nevertheless, there are those who advocate that the shrink-swell test is fundamentally flawed, generally on the basis that the reactivity index that results for any particular soil is affected by the water content of the soil at the time of measurement. Whilst there is anecdotal evidence to suggest this, the data usually put forward to demonstrate this assertion are seldom of high scientific quality, and not obtained under conditions that would allow the assertion to be rigorously and conclusively tested. In fact, scatter plots of shrink-swell index vs initial water content often put forward (eg Rogers and McDougall, 2020) as evidence do show a tendency for soils with higher water content to have higher shrink-swell index values, however, this paper demonstrates that this should actually be expected, and it does not, in itself, confirm that there is bias in the test. The paper discusses the alternative of basing foundation design on Atterberg limits, and recognises that whilst these give a good indication of the clay type and clay content of a soil, as they are determined on remoulded samples, they cannot account for soil structure which is a significant controlling factor in reactive soil behaviour. It goes on to consider what it would take to produce convincing data to demonstrate bias in the shrink-swell tests, and what would be needed to conclusively demonstrate that an alternative approach to site classification was more reliable.