Experimental evidence suggests that under certain circumstances the selectivity (or classification) function for hydrocyclones does not always conform the expected ’S’ shape but exhibits a dip in regions of the finer particle sizes (typically less than 20 mu m). This dip is known as the ’fish-hook effect’ and is of considerable industrial significance because it mediates the classification efficiency of ultra-fine materials. Only two previous attempts have been carried out to model the effect and, moreover, the existence of the phenomenon itself has been disputed. This paper presents a critical assessment of previous work and new evidence for its existence based on in situ particle size measurements on the process streams coupled with robust data reconciliation procedures. A theory to account for the phenomenon and a new empirical model are proposed and tested.