I'm very interested in the question of dogs learning by observation. The "decisive" paper that proved that dogs don't learn by observation was a very limited study, I believe, involving two dogs with no particular reinforcement history, in which the observer dog failed to learn some task. And somehow it went into the zeitgeist as a firm conclusion, one I was totally prepared to believe in myself, at one time. I thought that dolphins do learn by observation, and dogs don't. Then I started giving seminars, and seeing all kinds of observational learning going on.
Clicker trained dogs, at least, certainly do learn by observation of other dogs: the click tells the watcher-dog what the behaver-dog got paid for, so they try it. On one of my early websites there was a wonderful account of an Afghan learning obedience skills from its crate. While waiting for its turn at conformation training, it watched what its housemate golden, an obedience dog, was learning. It agitated in its crate, during the obedience class, until it was finally let out early, whereupon it grabbed a dumbbell and carried it over a jump. The owner then threw a little clicker training at it and got its CD in three weeks.
I often see my young poodle watch the old dog to see what she does, and then do the same. I now occasionally deliberately train one dog by tethering it to watch the other dog doing something new.
