The bias associated with the use of the shape of deformed inclusions to estimate bulk strain from ‘conglomerate’ systems is examined. Mechanical considerations based on Eshelby's theory for the deformation of ellipsoidal inclusions, supported by data from naturally deformed conglomerates with more than one type of inclusion, reveal a discrepancy between the degree of prolateness-oblateness shown by the inclusion and the bulk rock strain ellipsoid. This bias is most marked in situations where the difference of competence between the inclusion and matrix is high and the volume fraction of inclusions is low. In these situations, an inclusion which is more competent than the matrix will deform to a shape which is more prolate than the bulk strain ellipsoid. This has important implications for strain analysis of conglomeratic rocks and the interpretation of pebble shape fabrics in the context of regional tectonics.