Previous work on object naming has assumed that variation is mostly along a taxonomic axis, and in particular hierarchical. This parameter does not seem to explain the variation in ManyNames.

As shown in the table, in the vast majority of cases, no hierarchical relation between the name and the synset can be retrieved from WordNet. Even factoring in the noise introduced by referential uncertainty, it is clear that a good portion of our data cannot be explained by variation in the level of abstraction of the chosen name.