When using operant conditioning to train mice or rats, what type of stimulus is most effective? For example, does a negative reinforcement of a loud noise have a more profound effect than a negative reinforcement of pain? What about for positive reinforcement? Is pleasure more powerful than food? Which type of stimulus is considered optimal for learning?
$\begingroup$
$\endgroup$
-
1$\begingroup$ related: cogsci.stackexchange.com/questions/510/… and cogsci.stackexchange.com/questions/1092/… $\endgroup$ – Ofri Raviv Aug 18 '12 at 20:52
$\begingroup$
$\endgroup$
First, the concept of optimality of a learning curve is not well defined. You can measure at least 3 different aspects of learning:
- Speed of learning
- Time before extinction
- Performance at peak
Of course, there may be other measures as well, and any combination of such measure may also be a legitimate measure for certain uses.
Conditioned Taste Aversion (CTA) is a form of learning in which animals (and humans) learn to avoid a type of food after consuming it and suffering symptoms resembling food poisoning shortly after. In certain cases, the learning can take a single trial, cause the animal to completely avoid the food, and last for a lifetime [1], making this an optimal learning in all of the above 3 measures.
However, CTA is a very special and specific type of learning, and won't help you in teaching your rat to walk in a maze... I haven't found a methodological comparison of different types of reinforcements, and I suspect such a comparison might be very difficult or even impossible, because the efficacy of a (positive) reinforcement is greatly affected by the animal's "appetite" for it: for example, to use food as a reinforcement, rats are usually starved (the technical term is "maintained on a food deprivation schedule") for a few days. To use water as a reinforcement, rats are kept on a water deprivation schedule. How would you equate the subjective level of starvation and thirst?
Other aspects of the learning schedule have been studied more thoroughly. For example, to produce a long lasting learning (increase measure #2) at the expanse of speed of learning (measure #1) and peak performance (#3), you can provide partial reinforcement: reinforce the animal only on some of the trials [2].
[1] Conditioned taste aversion: Memory of a special kind. Bures, Jan; Bermúdez-Rattoni, Federico; Yamamoto, Takashi; Oxford University Press, 1998 doi: 10.1093/acprof:oso/9780198523475.001.0001
[2] Partial reinforcement: a review and critique. Jenkins, William O.; Stanley Jr., Julian C; Psychological Bulletin, Vol 47(3), May 1950 doi: 10.1037/h0060772
