JEP: ALC (Oct 2018)

So, this whole issue is about Pavlovian conditioning, and so I did not post anything about it. But then, something interesting happened. A new pre-print is out (nevermind my usual hatred of preprints, stick with me!). It shows that rats and ferrets learn more efficiently when they are NOT reinforced trial by trial. They argued that although reinforcement aids learning, is also masks knowledge! By this, they mean that reinforcement entrains certain responses (yes, S-R-O works!), but the problem is that the context of a test is not reflected in behavior. However, when you stop reinforcing, then context matters, and these animals seemingly KNOW a lot about what is going on.

So, I think this is an Emergent! [If you do not know what this means, you should tell David or I, because it means you need to read more of Duane Rumbaugh's work, especially if you ever intend to apply to be the Rumbaugh fellow.]

But, to our goal here, it made me remember a paper in this issue. Here it is:

The nature of phenotypic variation in Pavlovian conditioning. Pages 358-369. Iliescu, Adela F.; Hall, Jeremy; Wilkinson, Lawrence S.; Dwyer, Dominic M.; Honey, R. C.

And the abstract:

Pavlovian conditioning procedures result in dramatic individual differences in the topography of learnt behaviors in rats: When the temporary insertion of a lever into an operant chamber is paired with food pellets, some rats (known as sign-trackers) predominantly interact with the lever, while others (known as goal-trackers) predominantly approach the food well. Two experiments examined the sensitivity of these two behaviors to changing reinforcement contingencies in groups of male and female rats exhibiting the different phenotypes (i.e., sign-trackers and goal-trackers). In both phenotypes, behavior oriented to the food well was more sensitive to contingency changes (e.g., a reversal in which of two levers was reinforced) than was lever-oriented behavior. That is, the nature of the two behaviors differed independently of the rats in which they were manifest. These results indicate that the behavioral phenotypes reflect the parallel operation of a stimulus–stimulus associative process that gives rise to food-well activity and a stimulus–response process that gives rise to lever-oriented activity, rather than the operation of a single process (e.g., stimulus–stimulus) that generates both behaviors.

What this says, I think, is that if you look at each rat, it is not that some are sign trackers and some are goal-trackers, it is that they are all BOTH. They are S-S learners and S-R learners. Combine that with the preprint idea, that reinforcement masks actual knowledge that animals have, and then you can say they are S-S, S-R, and COGNITIVE creatures. In other words, all mechaniams are available, but selectively engaged, and context matters. Put another way, if you treat me like an associatively-trained creature, I act like one, but if you treat me like a knowledgeable cognitive agent, I act like one.

I think Duane would very much like these ideas! David Smith probably would too, although he would not be as accepting of the S-S and S-R features of the psychology of the individual, I think.