Bacteria and eukaryotes adopt very different strategies to modify their rRNAs. Most sites of eukaryotic rRNA modification are selected by guide small nucleolar RNAs (snoRNAs), while bacteria rely on numerous site-specific modification enzymes. This raises a 'chicken and egg' dilemma: how could a system of modification that requires a large number of snoRNA cofactors have developed? Did it arise in a de novo fashion, or evolve from a pre-existing protein-based system? The rRNA sequences are well conserved in evolution, but the pattern of modification is only moderately conserved, and many more sites are modified in eukaryotes than in bacteria; why is this so? We propose a model for the origins of the modification-guide snoRNAs that attempts to answer these questions.