On Scienceblogs.com, we find an article titled When crime-fighting tools go bad: Problems with the face-composite system that documents the difficulty that people have in describing faces.
The article here describes a problem with the use of software that helps police create sketches of perpetrators based on their victims’ verbal descriptions. The problem is that, when the victim only sees the perpetrator for a few seconds, but works with police on creating the software image for a much longer time, the victim remembers the sketch rather than the face of the actual perpetrator and selects people from police lineups accordingly.
This problem would not exist if natural language were up to the task of describing faces exactly, but apparently it isn’t.
Have very many studies been done on the difficulties that people have in verbally describing faces and other visual stimuli?