Although originally classified as galaxies, Ultra Compact Dwarfs (UCDs) share many properties in common with globular star clusters. The debate on the origin and nature of UCDs, and the recently discovered ultra-faint dwarf spheroidal (dSph) galaxies which contain very few stars, has motivated us to ask the question 'what is a galaxy?' Our aim here is to promote further discussion of how to define a galaxy and, in particular, what separates it from a star cluster. Like most previous definitions, we adopt the requirement of a gravitationally bound stellar system as a minimum. In order to separate a dwarf galaxy from a globular cluster, we discuss other possible requirements, such as a minimum size, a long two-body relaxation time, a satellite system, the presence of complex stellar populations and non-baryonic dark matter. We briefly mention the implications of each of these definitions if they are adopted. Some special cases of objects with an ambiguous nature are also discussed. Finally, we give our favoured criteria, and in the spirit of a 'collective wisdom', invite readers to vote on their preferred definition of a galaxy via a dedicated website.