Stephen Jay Gould did an article for his long running series where he argued that the intelligence of herbivores was actually a random walk.
If I understood and am remember it correctly, it went like this. For any particular species of herbivore, if you trace the cranial capacity of it's decendents, there is an equal chance of it having a higher and lower capacity. With the exception of course that there is a practical mininum necissary for the animal to function.
This sort of bounded random walk creates the illusion of a directed trend because the the highest value and the average value tend to go up over time.
Jay