Reappraisal of linear shrinkage test for plasticity index determination and classification of fine-grained soils

Among the multitude of alternate plasticity index (PI) estimation proposals, empirical correlations based on the linear shrinkage (LS) test, originally proposed in the 1967 edition of the British BS 1377 standard, seem to be gaining increased attention. This study has two main aims: ( i) re-examining LS-based correlations for PI estimation to better understand their true potentials and/or limitations; and ( ii) investigating the possibility of employing LS, as a proxy for the standard PI, for fine-grained soil classification. To this end, comprehensive statistical analyses were performed on the largest and most diverse database assembled of its kind, entailing PI:LS measurements for 265 different fine-grained soils [pertaining to the ‘soft’ base percussion-cup liquid limit (LL = 17–134%) and the Australian standard LS = 0.5–32.0%]. It was demonstrated that the LS can generally only provide a rough approximation of the actual PI. At best, only those predictions produced by PI = 1.86 × LS for low-plasticity soils (i.e., LL ≤ 35%) included in the database may be deemed acceptable, specifically when difficulties are encountered in executing the standard plastic limit (PL) thread-rolling test for such soils. The second/novel aim of this study found that the agreement level between the Casagrande-style plasticity-chart classification (based on the measured PI = LL − PL) and that performed using the PI deduced from a newly introduced multivariable correlation involving LS and LL, as inputs, was obtained as 95%. Hence, in the absence of measured PI, the multivariable PI:LS:LL correlation may be deemed suitable for preliminary soil classification purposes.