Aaron Blaisdell: UCLA Psychology and Brain Research Institute

I report a series of experiments showing that rats appear to make causal inferences in a basic task that taps
into core features of causal reasoning. (1) They derived predictions of the outcomes of interventions after
passive observational learning of different kinds of causal models. After learning through Pavlovian
observation that Event A was a common cause of Events X and Food (XßAàFood), rats predicted Food when
presented with Event X as a cue but discounted the alternative cause A when they generated X by means of a
lever press. (2) Rats showed evidence of reality monitoring. After learning an XàAàFood causal chain where
X was a tone and A was a light, when tested on X rats expected A to occur. But when A did not occur during
testing, rats did not expect food. By hiding the light during testing on X, however, rats showed no disruption of
food expectancy, suggesting that rats understood that A was unobservable. (3) Finally, we present evidence
that rats treated their actions as special in causal learning. Discounting of a previous cause was only
observed with interventions but not with other observable events; rats were capable of flexibly switching
between observational and interventional predictions; and discounting occurred on the very first test trial.
These results confirm causal-model theory but refute associative theories.
http://www.bec.ucla.edu/Blaisdell_Paper.pdf