A soil grading curve is one of the most useful tools to assess the suitability of material for rammed earth construction. Different and sometimes contradictory proportions of clay, silt, sand and gravel are proposed for rammed earth soils. This paper investigates the reliability of current guideline values through comparative performance testing of rammed earth specimens. Ten artificial soil batches (five of them stabilised with cement and/or lime) deemed suitable for rammed earth according to the current guidelines were tested in terms of compressive strength, shrinkage and erosion. The investigation shows that complying with soil particle size distribution criteria does not alone necessarily mean suitability of a soil for rammed earth. Based on these results, this paper proposes recommendations and criteria to be implemented in the assessment of soil for rammed earth.