I have seen a research project, where the subjects were presented with shapes to classify in two categories. The shapes were 3D models, some that were "pointy" and some that were "rounded".
The categories were labelled with two artificial words, one that subjects intuitively would classify as a "pointy" word, and the other a "rounded" word (only based on the configuration of consonants and vowels in the artificial word).
The effect is called the bouba/kiki effect. The article I seek used computer generated 3D models
Could you give a reference to this project? (Home page or citation).